{ "cells": [ { "cell_type": "code", "execution_count": 68, "metadata": {}, "outputs": [], "source": [ "import re\n", "import string\n", "from collections import Counter, defaultdict\n", "import spacy\n", "import textstat\n", "import sqlite3\n", "import json\n", "from tqdm import tqdm\n", "import pandas as pd\n", "from sklearn.preprocessing import StandardScaler\n", "import matplotlib.pyplot as plt\n", "from scipy.cluster.hierarchy import dendrogram, linkage, fcluster\n", "import numpy as np\n", "import random\n", "from pathlib import Path" ] }, { "cell_type": "code", "execution_count": 17, "metadata": {}, "outputs": [], "source": [ "# Load a small English model for spaCy\n", "nlp = spacy.load(\"en_core_web_sm\")\n", "\n", "# A list of common English function words.\n", "FUNCTION_WORDS = {\n", "\t\"the\", \"be\", \"to\", \"of\", \"and\", \"a\", \"in\", \"that\", \"have\", \"I\",\n", "\t\"for\", \"not\", \"on\", \"with\", \"he\", \"as\", \"you\", \"do\", \"at\", \"this\",\n", "\t\"but\", \"his\", \"by\", \"from\", \"they\", \"we\", \"say\", \"her\", \"she\",\n", "\t\"or\", \"an\", \"will\", \"my\", \"one\", \"all\", \"would\", \"there\", \"their\"\n", "}\n", "\n", "def syllable_count(word):\n", "\t\"\"\"Approximate the number of syllables in a word.\"\"\"\n", "\tword = word.lower()\n", "\t# A simple heuristic to estimate syllables.\n", "\t# There are more accurate methods, but this suffices for a rough measure.\n", "\tcount = 0\n", "\tvowels = 'aeiou'\n", "\t# Count vowel groups\n", "\tprevious_was_vowel = False\n", "\tfor char in word:\n", "\t\tif char in vowels:\n", "\t\t\tif not previous_was_vowel:\n", "\t\t\t\tcount += 1\n", "\t\t\tprevious_was_vowel = True\n", "\t\telse:\n", "\t\t\tprevious_was_vowel = False\n", "\t# Remove es and ed endings as extra syllables if they were counted\n", "\tif word.endswith('es') or word.endswith('ed'):\n", "\t\tcount = max(count - 1, 1)\n", "\treturn max(count, 1) # At least one syllable\n", "\n", "def get_char_bigrams(text):\n", "\tclean_text = re.sub(r'\\s+', ' ', text)\n", "\treturn [clean_text[i:i+2] for i in range(len(clean_text)-1)]" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "\"\"\"CREATE TABLE IF NOT EXISTS civitai (\n", "\tid INTEGER PRIMARY KEY,\n", "\turl TEXT NOT NULL,\n", "\twidth INTEGER NOT NULL,\n", "\theight INTEGER NOT NULL,\n", "\tnsfw BOOLEAN NOT NULL,\n", "\tnsfwLevel TEXT NOT NULL,\n", "\tcreatedAt TEXT NOT NULL,\n", "\tpostId INTEGER NOT NULL,\n", "\tcryCount INTEGER NOT NULL,\n", "\tlaughCount INTEGER NOT NULL,\n", "\tlikeCount INTEGER NOT NULL,\n", "\theartCount INTEGER NOT NULL,\n", "\tcommentCount INTEGER NOT NULL,\n", "\tmeta TEXT NOT NULL,\n", "\tusername TEXT,\n", "\timported BOOLEAN\n", ");\"\"\"\n", "civitai_conn = sqlite3.connect('civitai.sqlite')\n", "civitai_cursor = civitai_conn.cursor()\n", "civitai_cursor.execute('SELECT meta FROM civitai WHERE meta != \"null\"')\n", "\n", "unextracted = []\n", "all_prompts = set()\n", "\n", "for row in tqdm(civitai_cursor.fetchall()):\n", "\tmeta = row[0]\n", "\ttry:\n", "\t\tmeta = json.loads(meta)\n", "\texcept:\n", "\t\tunextracted.append(meta)\n", "\t\tcontinue\n", "\n", "\tif 'prompt' in meta:\n", "\t\tprompt = meta['prompt']\n", "\t\tall_prompts.add(prompt.strip())\n", "\n", "print(f\"Total prompts: {len(all_prompts)}\")\n", "print(f\"Unextracted: {len(unextracted)}\")" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "all_bigrams = Counter()\n", "for txt in all_prompts:\n", "\tbigrams = get_char_bigrams(txt)\n", "\tall_bigrams.update(bigrams)\n", "\n", "# Select the top N bigrams\n", "top_bigrams = [bg for bg, count in all_bigrams.most_common(100)]" ] }, { "cell_type": "code", "execution_count": 32, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "num_words : 26\n", "num_sentences : 3\n", "avg_word_length : 5.423076923076923\n", "word_length_variance : 8.653846153846155\n", "type_token_ratio : 0.9615384615384616\n", "avg_sentence_length : 8.666666666666666\n", "sentence_length_variance : 9.333333333333332\n", "function_word_frequency : 0.15384615384615385\n", "avg_syllables_per_word : 1.7692307692307692\n", "flesch_reading_ease : 62.64\n", "flesch_kincaid_grade : 6.7\n", "smog_index : 8.8\n", "gunning_fog : 8.1\n", "coleman_liau_index : 12.08\n", "automated_readability_index : 9.4\n", "dale_chall : 10.14\n", "pos_freq_ADJ : 0.07692307692307693\n", "pos_freq_ADP : 0.038461538461538464\n", "pos_freq_ADV : 0.0\n", "pos_freq_AUX : 0.11538461538461539\n", "pos_freq_CONJ : 0.0\n", "pos_freq_CCONJ : 0.038461538461538464\n", "pos_freq_DET : 0.038461538461538464\n", "pos_freq_INTJ : 0.0\n", "pos_freq_NOUN : 0.38461538461538464\n", "pos_freq_NUM : 0.0\n", "pos_freq_PART : 0.038461538461538464\n", "pos_freq_PRON : 0.07692307692307693\n", "pos_freq_PROPN : 0.038461538461538464\n", "pos_freq_PUNCT : 0.0\n", "pos_freq_SCONJ : 0.038461538461538464\n", "pos_freq_SYM : 0.0\n", "pos_freq_VERB : 0.11538461538461539\n", "pos_freq_X : 0.0\n", "punct_freq_! : 0.0\n", "punct_freq_\" : 0.0\n", "punct_freq_# : 0.0\n", "punct_freq_$ : 0.0\n", "punct_freq_% : 0.0\n", "punct_freq_& : 0.0\n", "punct_freq_' : 0.0\n", "punct_freq_( : 0.0\n", "punct_freq_) : 0.0\n", "punct_freq_* : 0.0\n", "punct_freq_+ : 0.0\n", "punct_freq_, : 0.07692307692307693\n", "punct_freq_- : 0.0\n", "punct_freq_. : 0.11538461538461539\n", "punct_freq_/ : 0.0\n", "punct_freq_: : 0.0\n", "punct_freq_; : 0.0\n", "punct_freq_< : 0.0\n", "punct_freq_= : 0.0\n", "punct_freq_> : 0.0\n", "punct_freq_? : 0.0\n", "punct_freq_@ : 0.0\n", "punct_freq_[ : 0.0\n", "punct_freq_\\ : 0.0\n", "punct_freq_] : 0.0\n", "punct_freq_^ : 0.0\n", "punct_freq__ : 0.0\n", "punct_freq_` : 0.0\n", "punct_freq_{ : 0.0\n", "punct_freq_| : 0.0\n", "punct_freq_} : 0.0\n", "punct_freq_~ : 0.0\n", "char_bigram_freq_, : 0.011764705882352941\n", "char_bigram_freq_in : 0.0\n", "char_bigram_freq_ s : 0.023529411764705882\n", "char_bigram_freq_e : 0.041176470588235294\n", "char_bigram_freq_re : 0.011764705882352941\n", "char_bigram_freq_ a : 0.011764705882352941\n", "char_bigram_freq_or : 0.0058823529411764705\n", "char_bigram_freq_ng : 0.0058823529411764705\n", "char_bigram_freq_er : 0.0\n", "char_bigram_freq_an : 0.011764705882352941\n", "char_bigram_freq_ b : 0.0058823529411764705\n", "char_bigram_freq_d : 0.011764705882352941\n", "char_bigram_freq_st : 0.023529411764705882\n", "char_bigram_freq_es : 0.0\n", "char_bigram_freq_s, : 0.0\n", "char_bigram_freq_th : 0.0058823529411764705\n", "char_bigram_freq_le : 0.01764705882352941\n", "char_bigram_freq_lo : 0.0\n", "char_bigram_freq_ t : 0.01764705882352941\n", "char_bigram_freq_he : 0.0\n", "char_bigram_freq_ti : 0.023529411764705882\n", "char_bigram_freq_s : 0.01764705882352941\n", "char_bigram_freq_ra : 0.011764705882352941\n", "char_bigram_freq_on : 0.01764705882352941\n", "char_bigram_freq_t : 0.011764705882352941\n", "char_bigram_freq_ c : 0.0\n", "char_bigram_freq_ h : 0.0058823529411764705\n", "char_bigram_freq_n : 0.0058823529411764705\n", "char_bigram_freq_ar : 0.0058823529411764705\n", "char_bigram_freq_g : 0.0\n", "char_bigram_freq_co : 0.0\n", "char_bigram_freq_at : 0.01764705882352941\n", "char_bigram_freq_nd : 0.0058823529411764705\n", "char_bigram_freq_ed : 0.0\n", "char_bigram_freq_al : 0.0\n", "char_bigram_freq_ea : 0.011764705882352941\n", "char_bigram_freq_li : 0.0058823529411764705\n", "char_bigram_freq_ai : 0.0\n", "char_bigram_freq_te : 0.023529411764705882\n", "char_bigram_freq_ f : 0.0058823529411764705\n", "char_bigram_freq_it : 0.0\n", "char_bigram_freq_ p : 0.0058823529411764705\n", "char_bigram_freq_e, : 0.0\n", "char_bigram_freq_ o : 0.0\n", "char_bigram_freq_de : 0.0058823529411764705\n", "char_bigram_freq_ d : 0.0058823529411764705\n", "char_bigram_freq_y : 0.011764705882352941\n", "char_bigram_freq_ic : 0.011764705882352941\n", "char_bigram_freq_ l : 0.0058823529411764705\n", "char_bigram_freq_en : 0.023529411764705882\n", "char_bigram_freq_ w : 0.0058823529411764705\n", "char_bigram_freq_ir : 0.0\n", "char_bigram_freq_ha : 0.0\n", "char_bigram_freq_sc : 0.0\n", "char_bigram_freq_ta : 0.0\n", "char_bigram_freq_hi : 0.0058823529411764705\n", "char_bigram_freq_et : 0.0058823529411764705\n", "char_bigram_freq_nt : 0.011764705882352941\n", "char_bigram_freq_ro : 0.0\n", "char_bigram_freq_ig : 0.0\n", "char_bigram_freq_ i : 0.011764705882352941\n", "char_bigram_freq_ac : 0.0058823529411764705\n", "char_bigram_freq_il : 0.0\n", "char_bigram_freq_ma : 0.0058823529411764705\n", "char_bigram_freq_e_ : 0.0\n", "char_bigram_freq_as : 0.0\n", "char_bigram_freq_ou : 0.0\n", "char_bigram_freq_ e : 0.011764705882352941\n", "char_bigram_freq_gh : 0.0\n", "char_bigram_freq_ m : 0.011764705882352941\n", "char_bigram_freq_is : 0.023529411764705882\n", "char_bigram_freq_a : 0.0058823529411764705\n", "char_bigram_freq_ce : 0.011764705882352941\n", "char_bigram_freq_ri : 0.0\n", "char_bigram_freq_r : 0.0\n", "char_bigram_freq_ne : 0.0\n", "char_bigram_freq_ur : 0.011764705882352941\n", "char_bigram_freq_l : 0.0\n", "char_bigram_freq_sh : 0.0\n", "char_bigram_freq_la : 0.0\n", "char_bigram_freq_t, : 0.0\n", "char_bigram_freq_h : 0.0\n", "char_bigram_freq_up : 0.0\n", "char_bigram_freq_ g : 0.0\n", "char_bigram_freq_se : 0.0058823529411764705\n", "char_bigram_freq_tr : 0.01764705882352941\n", "char_bigram_freq_ol : 0.0\n", "char_bigram_freq_r, : 0.0\n", "char_bigram_freq_ r : 0.0\n", "char_bigram_freq_ck : 0.0\n", "char_bigram_freq_wi : 0.0\n", "char_bigram_freq_of : 0.0\n", "char_bigram_freq_ho : 0.0058823529411764705\n", "char_bigram_freq_ve : 0.0\n", "char_bigram_freq_p, : 0.0\n", "char_bigram_freq_ow : 0.0058823529411764705\n", "char_bigram_freq_me : 0.0058823529411764705\n", "char_bigram_freq_ll : 0.0\n", "char_bigram_freq_bl : 0.0\n", "char_bigram_freq_to : 0.0058823529411764705\n" ] } ], "source": [ "def extract_stylistic_features(text, bigram_vocab=None):\n", "\t# Process text with spaCy\n", "\tdoc = nlp(text)\n", "\n", "\t# Extract tokens that are words (excluding punctuation, spaces, etc.)\n", "\twords = [token.text for token in doc if token.is_alpha]\n", "\tsentences = list(doc.sents)\n", "\tchars = list(text)\n", "\n", "\t# Basic counts\n", "\tnum_sentences = len(sentences) if len(sentences) > 0 else 1\n", "\tnum_words = len(words) if len(words) > 0 else 1\n", "\tnum_chars = sum(len(w) for w in words)\n", "\n", "\t# Lexical features\n", "\tword_lengths = [len(w) for w in words]\n", "\tavg_word_length = sum(word_lengths) / num_words if num_words > 0 else 0\n", "\tword_length_variance = 0.0\n", "\tif num_words > 1:\n", "\t\tmean_wl = avg_word_length\n", "\t\tword_length_variance = sum((wl - mean_wl)**2 for wl in word_lengths) / (num_words - 1)\n", "\n", "\t# Type-token ratio\n", "\tunique_words = set(w.lower() for w in words)\n", "\tttr = len(unique_words) / num_words if num_words > 0 else 0\n", "\n", "\t# Sentence length features\n", "\tsentence_lengths = [len([t for t in s if t.is_alpha]) for s in sentences]\n", "\tavg_sentence_length = sum(sentence_lengths)/num_sentences if num_sentences > 0 else 0\n", "\tsentence_length_variance = 0.0\n", "\tif num_sentences > 1:\n", "\t\tmean_sl = avg_sentence_length\n", "\t\tsentence_length_variance = sum((sl - mean_sl)**2 for sl in sentence_lengths) / (num_sentences - 1)\n", "\n", "\t# POS tag distribution\n", "\tpos_counts = Counter([token.pos_ for token in doc if token.is_alpha])\n", "\ttotal_pos = sum(pos_counts.values()) if sum(pos_counts.values()) > 0 else 1\n", "\n", "\t# Function word frequency\n", "\tlower_words = [w.lower() for w in words]\n", "\tfw_counts = Counter(lower_words)\n", "\ttotal_fw = sum(fw_counts[w] for w in FUNCTION_WORDS)\n", "\trelative_function_word_freq = total_fw / num_words if num_words > 0 else 0\n", "\n", "\t# Punctuation usage\n", "\tpunctuation_counts = Counter(ch for ch in text if ch in string.punctuation)\n", "\n", "\t# Syllable-based measures\n", "\tword_syllables = [syllable_count(w) for w in words]\n", "\tavg_syllables_per_word = sum(word_syllables)/num_words if num_words>0 else 0\n", "\n", "\t# Readability scores (using textstat)\n", "\tflesch_reading_ease = textstat.flesch_reading_ease(text)\n", "\tflesch_kincaid_grade = textstat.flesch_kincaid_grade(text)\n", "\tsmog_index = textstat.smog_index(text)\n", "\tgunning_fog = textstat.gunning_fog(text)\n", "\tcoleman_liau_index = textstat.coleman_liau_index(text)\n", "\tautomated_readability_index = textstat.automated_readability_index(text)\n", "\tdale_chall = textstat.dale_chall_readability_score(text)\n", "\n", "\t# Combine all features into a dictionary\n", "\tfeatures = {\n", "\t\t# Basic counts\n", "\t\t\"num_words\": num_words,\n", "\t\t\"num_sentences\": num_sentences,\n", "\t\t\n", "\t\t# Lexical\n", "\t\t\"avg_word_length\": avg_word_length,\n", "\t\t\"word_length_variance\": word_length_variance,\n", "\t\t\"type_token_ratio\": ttr,\n", "\t\t\n", "\t\t# Sentence\n", "\t\t\"avg_sentence_length\": avg_sentence_length,\n", "\t\t\"sentence_length_variance\": sentence_length_variance,\n", "\t\t\n", "\t\t# Function words\n", "\t\t\"function_word_frequency\": relative_function_word_freq,\n", "\t\t\n", "\t\t# Syllable measure\n", "\t\t\"avg_syllables_per_word\": avg_syllables_per_word,\n", "\t\t\n", "\t\t# Readability scores\n", "\t\t\"flesch_reading_ease\": flesch_reading_ease,\n", "\t\t\"flesch_kincaid_grade\": flesch_kincaid_grade,\n", "\t\t\"smog_index\": smog_index,\n", "\t\t\"gunning_fog\": gunning_fog,\n", "\t\t\"coleman_liau_index\": coleman_liau_index,\n", "\t\t\"automated_readability_index\": automated_readability_index,\n", "\t\t\"dale_chall\": dale_chall,\n", "\t}\n", "\n", "\t# POS tag distribution\n", "\tALL_POS_TAGS = [\n", "\t\t\"ADJ\", \"ADP\", \"ADV\", \"AUX\", \"CONJ\", \"CCONJ\", \"DET\", \"INTJ\", \"NOUN\", \"NUM\",\n", "\t\t\"PART\", \"PRON\", \"PROPN\", \"PUNCT\", \"SCONJ\", \"SYM\", \"VERB\", \"X\"\n", "\t]\n", "\n", "\tfor pos in ALL_POS_TAGS:\n", "\t\tcount = pos_counts.get(pos, 0)\n", "\t\tfeatures[f\"pos_freq_{pos}\"] = count / total_pos\n", "\n", "\t# Punctuation frequencies\n", "\tsafe_num_words = num_words if num_words > 0 else 1\n", "\tfor p in string.punctuation:\n", "\t\tcount = punctuation_counts[p] if p in punctuation_counts else 0\n", "\t\tfeatures[f\"punct_freq_{p}\"] = count / safe_num_words\n", "\n", "\t# Character-level features\n", " # Let's extract character bigrams\n", "\tchar_bigrams = get_char_bigrams(text)\n", "\tbigram_counts = Counter(char_bigrams)\n", "\ttotal_bigrams = sum(bigram_counts.values()) if bigram_counts else 1\n", "\n", "\tif bigram_vocab is not None:\n", "\t\tfor bg in bigram_vocab:\n", "\t\t\tfeatures[f\"char_bigram_freq_{bg}\"] = bigram_counts[bg] / total_bigrams\n", "\telse:\n", "\t\tpass\n", "\n", "\treturn features\n", "\n", "\n", "sample_text = \"\"\"This is a simple example text. It is meant to demonstrate stylistic feature extraction.\n", "Notice how punctuation, word length, and sentence structure may vary between texts.\"\"\"\n", "feats = extract_stylistic_features(sample_text, top_bigrams)\n", "for k, v in feats.items():\n", "\tprint(k, \":\", v)" ] }, { "cell_type": "code", "execution_count": 33, "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "100%|██████████| 69194/69194 [12:34<00:00, 91.76it/s] \n" ] } ], "source": [ "all_data = []\n", "\n", "for prompt in tqdm(all_prompts):\n", "\tfeats = extract_stylistic_features(prompt, top_bigrams)\n", "\tall_data.append({\n", "\t\t\"prompt\": prompt,\n", "\t\t\"features\": feats,\n", "\t})" ] }, { "cell_type": "code", "execution_count": 31, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "{'prompt': 'real, atmospheric scene, masterpiece, best quality, (detailed face, detail skin texture, ultra-detailed body),(cinematic light:1.1), 1girl, bttflorraine-smf, smile, realistic, grin, breasts, bare shoulders, cleavage, dress, black hair, solo focus, brown hair, strapless, medium breasts, lips, real life insert, pink dress ',\n", " 'features': {'num_words': 39,\n", " 'num_sentences': 1,\n", " 'avg_word_length': 5.897435897435898,\n", " 'word_length_variance': 5.3049932523616805,\n", " 'type_token_ratio': 0.8717948717948718,\n", " 'avg_sentence_length': 39.0,\n", " 'sentence_length_variance': 0.0,\n", " 'pos_freq_ADJ': 0.38461538461538464,\n", " 'pos_freq_NOUN': 0.5384615384615384,\n", " 'pos_freq_PROPN': 0.07692307692307693,\n", " 'function_word_frequency': 0.0,\n", " 'avg_syllables_per_word': 1.794871794871795,\n", " 'flesch_reading_ease': 25.8,\n", " 'flesch_kincaid_grade': 14.6,\n", " 'smog_index': 0.0,\n", " 'gunning_fog': 15.0,\n", " 'coleman_liau_index': 25.23,\n", " 'automated_readability_index': 27.4,\n", " 'dale_chall': 15.68,\n", " 'punct_freq_!': 0.0,\n", " 'punct_freq_\"': 0.0,\n", " 'punct_freq_#': 0.0,\n", " 'punct_freq_$': 0.0,\n", " 'punct_freq_%': 0.0,\n", " 'punct_freq_&': 0.0,\n", " \"punct_freq_'\": 0.0,\n", " 'punct_freq_(': 0.05128205128205128,\n", " 'punct_freq_)': 0.05128205128205128,\n", " 'punct_freq_*': 0.0,\n", " 'punct_freq_+': 0.0,\n", " 'punct_freq_,': 0.6153846153846154,\n", " 'punct_freq_-': 0.05128205128205128,\n", " 'punct_freq_.': 0.05128205128205128,\n", " 'punct_freq_/': 0.0,\n", " 'punct_freq_:': 0.07692307692307693,\n", " 'punct_freq_;': 0.0,\n", " 'punct_freq_<': 0.02564102564102564,\n", " 'punct_freq_=': 0.0,\n", " 'punct_freq_>': 0.02564102564102564,\n", " 'punct_freq_?': 0.0,\n", " 'punct_freq_@': 0.0,\n", " 'punct_freq_[': 0.0,\n", " 'punct_freq_\\\\': 0.0,\n", " 'punct_freq_]': 0.0,\n", " 'punct_freq_^': 0.0,\n", " 'punct_freq__': 0.10256410256410256,\n", " 'punct_freq_`': 0.0,\n", " 'punct_freq_{': 0.0,\n", " 'punct_freq_|': 0.0,\n", " 'punct_freq_}': 0.0,\n", " 'punct_freq_~': 0.0,\n", " 'char_bigram_freq_, ': 0.0625,\n", " 'char_bigram_freq_in': 0.019021739130434784,\n", " 'char_bigram_freq_ s': 0.016304347826086956,\n", " 'char_bigram_freq_e ': 0.005434782608695652,\n", " 'char_bigram_freq_re': 0.02717391304347826,\n", " 'char_bigram_freq_ a': 0.002717391304347826,\n", " 'char_bigram_freq_or': 0.010869565217391304,\n", " 'char_bigram_freq_ng': 0.0,\n", " 'char_bigram_freq_er': 0.010869565217391304,\n", " 'char_bigram_freq_an': 0.0,\n", " 'char_bigram_freq_ b': 0.021739130434782608,\n", " 'char_bigram_freq_d ': 0.005434782608695652,\n", " 'char_bigram_freq_st': 0.016304347826086956,\n", " 'char_bigram_freq_es': 0.010869565217391304,\n", " 'char_bigram_freq_s,': 0.019021739130434784,\n", " 'char_bigram_freq_th': 0.002717391304347826,\n", " 'char_bigram_freq_le': 0.01358695652173913,\n", " 'char_bigram_freq_lo': 0.01358695652173913,\n", " 'char_bigram_freq_ t': 0.002717391304347826,\n", " 'char_bigram_freq_he': 0.005434782608695652,\n", " 'char_bigram_freq_ti': 0.005434782608695652,\n", " 'char_bigram_freq_s ': 0.002717391304347826,\n", " 'char_bigram_freq_ra': 0.016304347826086956,\n", " 'char_bigram_freq_on': 0.0,\n", " 'char_bigram_freq_t ': 0.002717391304347826,\n", " 'char_bigram_freq_ c': 0.002717391304347826,\n", " 'char_bigram_freq_ h': 0.005434782608695652,\n", " 'char_bigram_freq_n ': 0.005434782608695652,\n", " 'char_bigram_freq_ar': 0.002717391304347826,\n", " 'char_bigram_freq_g ': 0.0,\n", " 'char_bigram_freq_co': 0.0,\n", " 'char_bigram_freq_at': 0.005434782608695652,\n", " 'char_bigram_freq_nd': 0.0,\n", " 'char_bigram_freq_ed': 0.008152173913043478,\n", " 'char_bigram_freq_al': 0.010869565217391304,\n", " 'char_bigram_freq_ea': 0.016304347826086956,\n", " 'char_bigram_freq_li': 0.01358695652173913,\n", " 'char_bigram_freq_ai': 0.019021739130434784,\n", " 'char_bigram_freq_te': 0.005434782608695652,\n", " 'char_bigram_freq_ f': 0.005434782608695652,\n", " 'char_bigram_freq_it': 0.002717391304347826,\n", " 'char_bigram_freq_ p': 0.002717391304347826,\n", " 'char_bigram_freq_e,': 0.016304347826086956,\n", " 'char_bigram_freq_ o': 0.0,\n", " 'char_bigram_freq_de': 0.010869565217391304,\n", " 'char_bigram_freq_ d': 0.008152173913043478,\n", " 'char_bigram_freq_y ': 0.0,\n", " 'char_bigram_freq_ic': 0.008152173913043478,\n", " 'char_bigram_freq_ l': 0.008152173913043478,\n", " 'char_bigram_freq_en': 0.002717391304347826,\n", " 'char_bigram_freq_ w': 0.0,\n", " 'char_bigram_freq_ir': 0.008152173913043478,\n", " 'char_bigram_freq_ha': 0.005434782608695652,\n", " 'char_bigram_freq_sc': 0.002717391304347826,\n", " 'char_bigram_freq_ta': 0.008152173913043478,\n", " 'char_bigram_freq_hi': 0.0,\n", " 'char_bigram_freq_et': 0.008152173913043478,\n", " 'char_bigram_freq_nt': 0.0,\n", " 'char_bigram_freq_ro': 0.002717391304347826,\n", " 'char_bigram_freq_ig': 0.002717391304347826,\n", " 'char_bigram_freq_ i': 0.002717391304347826,\n", " 'char_bigram_freq_ac': 0.008152173913043478,\n", " 'char_bigram_freq_il': 0.010869565217391304,\n", " 'char_bigram_freq_ma': 0.005434782608695652,\n", " 'char_bigram_freq_e_': 0.005434782608695652,\n", " 'char_bigram_freq_as': 0.008152173913043478,\n", " 'char_bigram_freq_ou': 0.002717391304347826,\n", " 'char_bigram_freq_ e': 0.0,\n", " 'char_bigram_freq_gh': 0.002717391304347826,\n", " 'char_bigram_freq_ m': 0.005434782608695652,\n", " 'char_bigram_freq_is': 0.002717391304347826,\n", " 'char_bigram_freq_a ': 0.0,\n", " 'char_bigram_freq_ce': 0.008152173913043478,\n", " 'char_bigram_freq_ri': 0.005434782608695652,\n", " 'char_bigram_freq_r ': 0.0,\n", " 'char_bigram_freq_ne': 0.010869565217391304,\n", " 'char_bigram_freq_ur': 0.005434782608695652,\n", " 'char_bigram_freq_l ': 0.005434782608695652,\n", " 'char_bigram_freq_sh': 0.002717391304347826,\n", " 'char_bigram_freq_la': 0.002717391304347826,\n", " 'char_bigram_freq_t,': 0.002717391304347826,\n", " 'char_bigram_freq_h ': 0.0,\n", " 'char_bigram_freq_up': 0.0,\n", " 'char_bigram_freq_ g': 0.002717391304347826,\n", " 'char_bigram_freq_se': 0.002717391304347826,\n", " 'char_bigram_freq_tr': 0.005434782608695652,\n", " 'char_bigram_freq_ol': 0.002717391304347826,\n", " 'char_bigram_freq_r,': 0.005434782608695652,\n", " 'char_bigram_freq_ r': 0.005434782608695652,\n", " 'char_bigram_freq_ck': 0.005434782608695652,\n", " 'char_bigram_freq_wi': 0.0,\n", " 'char_bigram_freq_of': 0.0,\n", " 'char_bigram_freq_ho': 0.002717391304347826,\n", " 'char_bigram_freq_ve': 0.0,\n", " 'char_bigram_freq_p,': 0.0,\n", " 'char_bigram_freq_ow': 0.002717391304347826,\n", " 'char_bigram_freq_me': 0.002717391304347826,\n", " 'char_bigram_freq_ll': 0.0,\n", " 'char_bigram_freq_bl': 0.002717391304347826,\n", " 'char_bigram_freq_to': 0.002717391304347826}}" ] }, "execution_count": 31, "metadata": {}, "output_type": "execute_result" } ], "source": [ "all_data[0]" ] }, { "cell_type": "code", "execution_count": 34, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "(69194, 166)\n" ] }, { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
num_wordsnum_sentencesavg_word_lengthword_length_variancetype_token_ratioavg_sentence_lengthsentence_length_variancefunction_word_frequencyavg_syllables_per_wordflesch_reading_ease...char_bigram_freq_wichar_bigram_freq_ofchar_bigram_freq_hochar_bigram_freq_vechar_bigram_freq_p,char_bigram_freq_owchar_bigram_freq_mechar_bigram_freq_llchar_bigram_freq_blchar_bigram_freq_to
0-0.180893-0.2834040.771401-0.0431420.2490950.091864-0.027033-1.1013150.461159-0.123350...-0.695055-0.684310-0.134455-0.802422-0.642685-0.044840-0.077319-0.633590-0.072572-0.008556
1-0.562361-0.283404-8.229486-1.386490-7.465739-1.236617-0.027033-1.101315-7.3296983.177128...-0.695055-0.684310-0.787097-0.802422-0.642685-0.690976-0.681372-0.633590-0.768923-0.724235
2-0.351550-0.2834041.4135590.2627500.981385-0.487217-0.027033-0.7193270.562338-0.296177...-0.6950551.598209-0.787097-0.802422-0.6426850.4861431.519535-0.633590-0.768923-0.724235
3-0.371627-0.2834041.080562-0.1363660.941161-0.555345-0.027033-1.1013150.4834180.178914...-0.695055-0.684310-0.787097-0.8024220.934763-0.6909760.447012-0.633590-0.768923-0.724235
40.039957-0.2834040.077249-0.4029320.3681280.841263-0.027033-0.136952-0.071612-0.394622...-0.223185-0.251788-0.3364920.2792050.231867-0.244863-0.264314-0.2687400.192642-0.724235
..................................................................
691890.371231-0.283404-0.387224-0.250519-0.0285041.965363-0.027033-0.117899-0.8649450.364685...-0.383012-0.3982870.1068450.2704820.513980-0.395966-0.1297800.572768-0.450986-0.397474
69190-0.301356-0.020266-1.389681-0.6975200.400366-0.776758-0.026935-0.167567-1.0599130.903765...-0.695055-0.684310-0.787097-0.802422-0.6426852.226546-0.6813721.752484-0.768923-0.724235
69191-0.110623-0.2834041.2929080.450052-0.5401470.330309-0.027033-0.0051760.691013-0.379491...-0.6950550.390443-0.7870971.213338-0.642685-0.136715-0.681372-0.180292-0.768923-0.110319
691920.120266-0.2834040.109521-0.495891-1.8226641.113772-0.027033-0.979522-0.032424-0.728061...-0.695055-0.6843102.1215650.1949950.163779-0.2795950.4723910.0392990.561131-0.724235
69193-0.281279-0.2834041.085825-0.3130970.773328-0.248772-0.027033-1.1013150.453483-0.135747...-0.6950550.3635740.304596-0.802422-0.642685-0.6909760.329044-0.6335901.560687-0.724235
\n", "

69194 rows × 166 columns

\n", "
" ], "text/plain": [ " num_words num_sentences avg_word_length word_length_variance \\\n", "0 -0.180893 -0.283404 0.771401 -0.043142 \n", "1 -0.562361 -0.283404 -8.229486 -1.386490 \n", "2 -0.351550 -0.283404 1.413559 0.262750 \n", "3 -0.371627 -0.283404 1.080562 -0.136366 \n", "4 0.039957 -0.283404 0.077249 -0.402932 \n", "... ... ... ... ... \n", "69189 0.371231 -0.283404 -0.387224 -0.250519 \n", "69190 -0.301356 -0.020266 -1.389681 -0.697520 \n", "69191 -0.110623 -0.283404 1.292908 0.450052 \n", "69192 0.120266 -0.283404 0.109521 -0.495891 \n", "69193 -0.281279 -0.283404 1.085825 -0.313097 \n", "\n", " type_token_ratio avg_sentence_length sentence_length_variance \\\n", "0 0.249095 0.091864 -0.027033 \n", "1 -7.465739 -1.236617 -0.027033 \n", "2 0.981385 -0.487217 -0.027033 \n", "3 0.941161 -0.555345 -0.027033 \n", "4 0.368128 0.841263 -0.027033 \n", "... ... ... ... \n", "69189 -0.028504 1.965363 -0.027033 \n", "69190 0.400366 -0.776758 -0.026935 \n", "69191 -0.540147 0.330309 -0.027033 \n", "69192 -1.822664 1.113772 -0.027033 \n", "69193 0.773328 -0.248772 -0.027033 \n", "\n", " function_word_frequency avg_syllables_per_word flesch_reading_ease \\\n", "0 -1.101315 0.461159 -0.123350 \n", "1 -1.101315 -7.329698 3.177128 \n", "2 -0.719327 0.562338 -0.296177 \n", "3 -1.101315 0.483418 0.178914 \n", "4 -0.136952 -0.071612 -0.394622 \n", "... ... ... ... \n", "69189 -0.117899 -0.864945 0.364685 \n", "69190 -0.167567 -1.059913 0.903765 \n", "69191 -0.005176 0.691013 -0.379491 \n", "69192 -0.979522 -0.032424 -0.728061 \n", "69193 -1.101315 0.453483 -0.135747 \n", "\n", " ... char_bigram_freq_wi char_bigram_freq_of char_bigram_freq_ho \\\n", "0 ... -0.695055 -0.684310 -0.134455 \n", "1 ... -0.695055 -0.684310 -0.787097 \n", "2 ... -0.695055 1.598209 -0.787097 \n", "3 ... -0.695055 -0.684310 -0.787097 \n", "4 ... -0.223185 -0.251788 -0.336492 \n", "... ... ... ... ... \n", "69189 ... -0.383012 -0.398287 0.106845 \n", "69190 ... -0.695055 -0.684310 -0.787097 \n", "69191 ... -0.695055 0.390443 -0.787097 \n", "69192 ... -0.695055 -0.684310 2.121565 \n", "69193 ... -0.695055 0.363574 0.304596 \n", "\n", " char_bigram_freq_ve char_bigram_freq_p, char_bigram_freq_ow \\\n", "0 -0.802422 -0.642685 -0.044840 \n", "1 -0.802422 -0.642685 -0.690976 \n", "2 -0.802422 -0.642685 0.486143 \n", "3 -0.802422 0.934763 -0.690976 \n", "4 0.279205 0.231867 -0.244863 \n", "... ... ... ... \n", "69189 0.270482 0.513980 -0.395966 \n", "69190 -0.802422 -0.642685 2.226546 \n", "69191 1.213338 -0.642685 -0.136715 \n", "69192 0.194995 0.163779 -0.279595 \n", "69193 -0.802422 -0.642685 -0.690976 \n", "\n", " char_bigram_freq_me char_bigram_freq_ll char_bigram_freq_bl \\\n", "0 -0.077319 -0.633590 -0.072572 \n", "1 -0.681372 -0.633590 -0.768923 \n", "2 1.519535 -0.633590 -0.768923 \n", "3 0.447012 -0.633590 -0.768923 \n", "4 -0.264314 -0.268740 0.192642 \n", "... ... ... ... \n", "69189 -0.129780 0.572768 -0.450986 \n", "69190 -0.681372 1.752484 -0.768923 \n", "69191 -0.681372 -0.180292 -0.768923 \n", "69192 0.472391 0.039299 0.561131 \n", "69193 0.329044 -0.633590 1.560687 \n", "\n", " char_bigram_freq_to \n", "0 -0.008556 \n", "1 -0.724235 \n", "2 -0.724235 \n", "3 -0.724235 \n", "4 -0.724235 \n", "... ... \n", "69189 -0.397474 \n", "69190 -0.724235 \n", "69191 -0.110319 \n", "69192 -0.724235 \n", "69193 -0.724235 \n", "\n", "[69194 rows x 166 columns]" ] }, "execution_count": 34, "metadata": {}, "output_type": "execute_result" } ], "source": [ "df = pd.DataFrame([item[\"features\"] for item in all_data])\n", "\n", "scaler = StandardScaler()\n", "scaled_features = scaler.fit_transform(df)\n", "print(scaled_features.shape)\n", "\n", "scaled_df = pd.DataFrame(scaled_features, columns=df.columns)\n", "scaled_df" ] }, { "cell_type": "code", "execution_count": 53, "metadata": {}, "outputs": [ { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "# Perform hierarchical clustering using Ward's method (commonly used)\n", "#foo = np.random.choice(scaled_features.shape[0], 16000, replace=False)\n", "Z = linkage(scaled_features, method='ward')\n", "\n", "# Plot the dendrogram\n", "plt.figure(figsize=(12, 8))\n", "dendrogram(Z)\n", "plt.title(\"Hierarchical Clustering Dendrogram\")\n", "plt.xlabel(\"Samples\")\n", "plt.ylabel(\"Distance\")\n", "plt.show()" ] }, { "cell_type": "code", "execution_count": 65, "metadata": {}, "outputs": [], "source": [ "clusters = fcluster(Z, t=300, criterion='distance')" ] }, { "cell_type": "code", "execution_count": 67, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Cluster 5 (9406 prompts):\n", " real, atmospheric scene, masterpiece, best quality...\n", " pixar style of walrus, as a pixar character, c...\n", " sinister, eerie, dark, shadow, distorted, grainy, ...\n", " 1woman, solo, long hair, pink hair, black choker,...\n", " orange peel, bright lights, ultra realistic, space...\n", "\n", "Cluster 11 (9276 prompts):\n", " ...\n", " Medium Otherworldly Anomalous Textured Arachnid, ...\n", " score_9, score_8_up, score_7_up, score_6_up, sourc...\n", " from front,1girl,random fictional girl named Maya ...\n", " masterpiece, best quality, illustration, 1boy, sol...\n", "\n", "Cluster 19 (7157 prompts):\n", " score_9, score_8_up, score_7_up, best quality, ro...\n", " score_9, score_8_up, score_7_up, score_6_up, score...\n", " (score_9, score_8_up, score_7_up, score_6_up, scor...\n", " score_9, score_8_up, score_7_up, BREAK\n", "(first pers...\n", " safe_pos, Expressiveh, score_9, score_8_up, score_...\n", "\n", "Cluster 21 (8946 prompts):\n", " (score_9, score_8_up, score_7_up), 1girl, white ha...\n", " safe_pos, Warcraft style, masterpiece, score_9, sc...\n", " score_9,score_8_up,core_7_up, 1girl, foot up...\n", " ,by Hugh ...\n", " z03-001 epiCRealism ...\n", " by (((Natalia Rak) and Jeremiah Ketner) and Daniel...\n", " \u0004B\u0004@\u0004O\u0004A\u00040\u0004N\u0004I\u00045 \u0004:\u0004@\u00040\u0004A\u0004>\u0004G...\n", "\n", "Cluster 18 (1 prompts):\n", " masterpiece,(bestquality),highlydetailed,ultra-det...\n", "\n", "Cluster 12 (2 prompts):\n", " (Highest picture quality),(Master's work),(ultra-d...\n", " Conceptual Art of (Clumsy Drawing:1.3) of (Cel sha...\n", "\n", "Cluster 9 (5 prompts):\n", " (((The most amazing scene ever created)))+++, (((F...\n", " +...\n", " tensor.art flux dev+Lora Hopper+upscaler...\n", " dfgvnrvnklsfvslkfnvaeñolfvnmañldskjbvnalñkva\n", "fire\n", "...\n", " tensor.art flux dev +Lora+upscaler...\n", "\n", "Cluster 13 (1 prompts):\n", " {\n", " \"W\": 768,\n", " \"H\": 1216,\n", " \"show_info_on_u...\n", "\n", "Cluster 25 (1 prompts):\n", " _...\n", " o_.-\"` `\\\n", " .--. _ `'-...\n", "\n", "Cluster 14 (1 prompts):\n", " 1, A vibrant constructivist rendering of New York ...\n", "\n" ] } ], "source": [ "by_cluster = defaultdict(list)\n", "#for i, idx in enumerate(foo):\n", "for idx in range(len(all_data)):\n", "\tcluster = clusters[idx]\n", "\tprompt = all_data[idx]['prompt']\n", "\tby_cluster[cluster].append(prompt)\n", "\n", "for cluster, prompts in by_cluster.items():\n", "\tprint(f\"Cluster {cluster} ({len(prompts)} prompts):\")\n", "\tfor p in prompts[:5]:\n", "\t\tprint(f\" {p[:50]}...\")\n", "\tprint()" ] }, { "cell_type": "code", "execution_count": 70, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "31756622" ] }, "execution_count": 70, "metadata": {}, "output_type": "execute_result" } ], "source": [ "Path(\"prompts_by_cluster.json\").write_text(json.dumps({str(k): v for k, v in by_cluster.items()}, indent=2))" ] }, { "cell_type": "code", "execution_count": 78, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "A movie poster for 'A Zoom Interview with a Vampire.' Include the zoom logo.\n", "A movie poster for 'A Zoom Interview with a Vampire.' Include the zoom logo.\n", "A movie poster for 'A Zoom Interview with a Vampire.' (Include the zoom logo.\n" ] } ], "source": [ "def remove_loras(prompt: str) -> str:\n", "\t# Remove LoRA tags of the form \n", "\treturn re.sub(r\"]+>\", \"\", prompt)\n", "\n", "def remove_scaling_factors(text: str) -> str:\n", "\t# Remove trailing scaling factors like :1.2, :0.5, etc.\n", "\t# This matches a colon followed by a number (possibly with a decimal) at the end of the string\n", "\treturn re.sub(r\":\\s*\\d+(\\.\\d+)?\\s*$\", \"\", text.strip())\n", "\n", "def strip_brackets(prompt: str) -> str:\n", "\tresult = \"\"\n", "\tstack = []\n", "\n", "\tfor ch in prompt:\n", "\t\tif ch in \"([\":\n", "\t\t\tstack.append(ch)\n", "\t\telif len(stack) > 0 and ((stack[-1][0] == \"(\" and ch == \")\") or (stack[-1][0] == \"[\" and ch == \"]\")):\n", "\t\t\tinner = stack.pop()[1:]\n", "\t\t\t# Remove scaling factors from inner\n", "\t\t\tinner = remove_scaling_factors(inner)\n", "\t\t\t\n", "\t\t\tif len(stack) > 0:\n", "\t\t\t\tstack[-1] += inner\n", "\t\t\telse:\n", "\t\t\t\tresult += inner\n", "\t\telif len(stack) > 0:\n", "\t\t\tstack[-1] += ch\n", "\t\telse:\n", "\t\t\tresult += ch\n", "\t\n", "\t# Handle any dangling brackets\n", "\twhile len(stack) > 0:\n", "\t\tinner = stack.pop()\n", "\t\tif len(stack) > 0:\n", "\t\t\tstack[-1] += inner\n", "\t\telse:\n", "\t\t\tresult += inner\n", "\n", "\treturn result\n", "\n", "def clean_prompt(prompt: str) -> str:\n", "\tprompt = remove_loras(prompt)\n", "\tprompt = strip_brackets(prompt)\n", "\t# Cleanup extra whitespace\n", "\t#prompt = re.sub(r\"\\s+\", \" \", prompt).strip()\n", "\treturn prompt\n", "\n", "# Test the given prompt\n", "test_prompt = \"(((A movie poster for 'A Zoom Interview with a Vampire.')) Include the zoom logo. )\"\n", "print(clean_prompt(test_prompt))\n", "test_prompt = \"(A movie:1.2) poster for 'A Zoom Interview with a Vampire.' Include the zoom logo.\"\n", "print(clean_prompt(test_prompt))\n", "test_prompt = \"(A movie : 1.8 ) poster for 'A Zoom Interview with a Vampire.' (Include the zoom logo.\"\n", "print(clean_prompt(test_prompt))" ] }, { "cell_type": "code", "execution_count": 80, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Cluster 1:\n", "```\n", "[\n", " \"A tiny mouse stands on a colorful pile of fresh fruits in a bustling outdoor market. The mouse holds a small, red tomato with both paws, nibbling it eagerly. Sunlight filters through the scene, highlighting the vibrant colors of oranges, limes, and tomatoes around it. The mouse\\u2019s big eyes look up occasionally, wide with excitement and curiosity. Background figures of market-goers pass by in a blur, emphasizing the mouse\\u2019s small size and playful, sneaky demeanor as it enjoys its snack undisturbed\",\n", " \"A 45 years old slightly overweight woman stands in front of a mirror, capturing a selfie. The image quality is grainy, with a slight blur softening the details. The lighting is dim, casting shadows that obscure her features. The room is cluttered, with clothes strewn across the bed and an unmade blanket. Her expression is casual, full of concentration, while the old iPhone struggles to focus, giving the photo an authentic, unpolished feel. The mirror shows smudges and fingerprints, adding to the raw, everyday atmosphere of the scene.\",\n", " \"Silhouetted figures, their forms reduced to powerful, detailed shapes, stand in contrast to ethereal, light-filled backgrounds. In a surreal landscape, the brains of a man and woman twist and dance, swirling patterns of emotion and energy pulsing through the scene. An air of melancholy and mystery pervades the minimalist, yet deeply evocative, composition\",\n", " \"In the heart of a gothic cathedral, shrouded in an ethereal darkness, a breathtakingly elegant woman, her form fragmented and encased within a majestic stained glass raven, reigns over a kingdom of eternal night. The raven, a symphony of shimmering blacks, deep blues, and iridescent purples, dominates the scene, its intricate stained glass feathers swirling around the woman like shadows given life. Her long, flowing black hair, interwoven with strands of obsidian and amethyst, cascades down her shoulders, blending seamlessly with the raven's dark plumage. \\n\\nHer fragmented form, a masterpiece of stained glass artistry, is composed of vibrant panels depicting stylized raven feathers, swirling patterns, and celestial constellations, each piece outlined in delicate lead lines of blackened silver. Her crystal dress is transformed into a shimmering cloak of stained glass feathers, echoing the raven's enigmatic form and glowing softly with an inner light in shades of deep sapphire and amethyst. An elaborate silver mask, adorned with stained glass accents in blacks, blues, and purples, obscures her face, adding an air of mystery and otherworldly allure.\\n\\nThe cathedral, a gothic masterpiece bathed in the cool light of a hidden moon, is adorned with massive stained glass windows depicting nocturnal scenes, swirling patterns, and celestial bodies in shades of midnight blue, obsidian black, and deep violet. The woman is surrounded by a profusion of dark, exotic flowers, their velvety petals and shimmering leaves creating a rich tapestry of textures and colors that complement the stained glass aesthetic. \\n\\nThe overall effect is one of haunting beauty and enigmatic power, a mesmerizing tableau that evokes a sense of gothic mystery and the eternal secrets of the night. The woman, a raven trapped within her stained glass prison, becomes a symbol of both knowledge and shadow, a testament to the enduring power of art and the timeless allure of the unknown. Rendered in hyperrealistic detail with a soft focus and a stained glass effect, the image captures the intricate details of the raven, the woman's fragmented form, the dark flora, and the gothic architecture with breathtaking clarity, inviting the viewer to lose themselves in this mesmerizing world of stained glass and shadows, where ancient secrets and hidden knowledge await.\",\n", " \"Darkside of the moon, moon Style By Duncan Jones, Capture the juxtaposition of design and artistry: a Spaceship standing before a painted Neon Space Artpyrmdmrs Pyramide wall. The mural, awash in colorful shades, seamlessly extends the tree's branches, blurring the lines between reality and imagination. Show how distance transforms this scene, revealing the harmony between the natural world and human creativity.\",\n", " \"A Powerful BlackT40k space marine, draped in tattered, dark power armor that pulses with a faint, eerie glow, stands on a creaking wooden bridge. He points the tip of a massive, rune-etched sword menacingly at the camera, the blade crackling with dark energy. Shadows swirl around him, and the misty air is thick with tension. red flames flicker within his hollow eyes, reflecting the bridge's warped, splintered planks, over the sword camera view \",\n", " \"a realistic picture,Display the text \\\"i sell fish for buzz!!\\\" on top a movie poster\\nMovie poster of 2 orange cats wearing black military tactical pants and military boots and wearing black hoodie, and tactical gears, the orange cats standing in front fish stall in market, image of a hand offering BUZZ COIN,the usual stacks of cash Instead, the shelves are lined with small, neatly arranged coins, each shaped like a tiny, stylized orange lightning bolt. The coins are labeled as \\\"Buzz.\\\") to a fishmonger cat, holding fish, 8k, ultra detailed, at the market, I background: traditional fish market ,his two paws are grasping onto a bright yellow buzz coin,that is almost his size as he is trying to drag it off the counter top, logo:1.9 , full body shots,\\nDisplay the text \\\"i sell fish for buzz!!\\\" on a movie poster\",\n", " \"A cute curious, small alien creature peers out from behind a lush fern in the dense, vibrant jungle, investigating where a spaceship wreckage lies half-submerged in the murky water. A distress beacon on the wreckage displays a warning message: 'Low Buzz'.\"\n", "]\n", "```\n", "\n", "Cluster 2:\n", "```\n", "[\n", " \"Black and white, doodle in notebook style of A stormtrooper in a yellow dress is standing next to a stack of suitcases. He is wearing a hat and sunglasses and is holding a checklist. The background is white. The style is simple and graphic. The stormtrooper is drawn in a cartoon style, 32k resolution, molecular precision, ultra sharp focus\",\n", " \"score_9, score_8_up, score_7_up, score_6_up, score_5_up, shawl, holding flower, !, sharp teeth, tarot, thighhighs, black hair, nature, 1girl 4t4r\",\n", " \"featuring a dynamic arrangement of potions and scrolls, diverse fantasy geometric shapes and vivid colors. The public will be able to explore their own lives and crafts, developing a new perspective on culture, creating new realities, new connections, meanings and transforming the world around them. Thus, the current artistic expression in the form of contemporary dance appears to be preeminent. by David A Hardy, media,grass,fence,road,house,power lines,utility pole,lake,pagoda, cloud,town,chimney,nature,stairs,bridge,river,pier,barefoot,fishing rod,dock,real world location,ship,mountain,fishing rod,fishing, inlay myanmar\",\n", " \"comic best quality, masterpiece, intricate, steelheartQuiron, Dancing with arms swinging side to side . graphic illustration, comic art, graphic novel art, vibrant, highly detailed\",\n", " \"Armarouge, attempting to cook marshmallows on his own fiery ponytail\",\n", " \"toad from supermario with a pampmpkn head made out of pumpkin. he is wearing a blue vest and a white shorts. he is in a super mario level with autumn decoration in the background and pumpkins. the picture gives a cozy feeling\",\n", " \"The image is a 3D rendering of a Bruce Lee made up of broccoli. He is in a fighting stance, with his arms stretched out to the sides and his legs bent at the knees. He is holding a bunch of broccoli in his right hand and his left hand is pointing towards the right side of the image. The broccoli is a bright green color and appears to be fresh and healthy. The background is a matte gradient of colorful gray, making him and the broccoli stand out.\",\n", " \"hyperrealistic portrait of a fierce young woman with pigtails, intense gaze, Union Jack top, detailed background with a chaotic room filled with sports equipment and graffiti, soft sunlight streaming through dusty windows, cinematic, 8k, ultra high detail, trending on artstation\"\n", "]\n", "```\n", "\n", "Cluster 3:\n", "```\n", "[\n", " \"AnjelikaV2, \\nAbstract circuit board, glowing blue lines, and a futuristic tech design, Geometric pattern, dark background with neon highlights in multiple colors, Dark fantasy character, 1girl, intricate costume with dark and light elements, high detail, and a moody, mysterious atmosphere.\\nStanding with one arm wrapped around the waist, hand on opposite shoulder\",\n", " \"Voluptuous Attractive, 30 year old woman, straight auburn hair, slutty, pouting, big eyes, close-up \\nOutside\\n\",\n", " \"XUER guangying,outdoors,sky,day,cloud,water,tree,building,scenery,aircraft,watercraft,bridge,boat\\uff0c\\n\\ns of a Sci-Fi Cityscape. This is a highly detailed, digital illustration depicting a futuristic cityscape set against a vibrant blue sky with wispy clouds. The city, with its sleek, organic architecture, is characterized by numerous towering, futuristic structures that resemble abstract sculptures with smooth, flowing curves and intricate patterns. The structures are predominantly white and silver, with some sections featuring translucent panels that emit a soft, blue glow. \\n\\nIn the foreground, the calm, turquoise waters of a lake or bay are prominently featured, with a white sailboat gracefully gliding on the water. The sailboat adds a sense of tranquility and exploration to the scene. \\n\\nThe background showcases a vast, alien landscape with multiple, large, gas giant planets, their swirling, colorful atmospheres adding a surreal touch. Two smaller, blue planets are also visible, floating in the distance. The cityscape is adorned with numerous, slender, pointed towers that rise majestically into the sky, adding to the sense of grandeur and futuristic ambiance. \\n\\nThe overall color palette is dominated by cool tones, with the exception of the sailboat, which stands out in its stark white against the blue and gray backdrop. The illustration is rich in detail, capturing the essence of a utopian, futuristic society.\\n\\nvery aesthetic, best quality,ultra detailed,intricate details,dark theme,soothing tones,high contrast,natural skin texture,soft light,sharp,masterpiece,best quality,ray tracing,sunlight,scenic view,vines,invasive flowers,with a high-end texture,in the style of fashion photography,magazine cover,Dynamic Angle,Dynamic posture,Perspective,High Point,pov,8k, RAW photo, highly detailed,masterpiece, highest quality,rich colors,high contrast,full shot body photo of the most beautiful artwork in the world,cinematic light,fantasy,highres,detailed face,ultra-realistic,high dynamic range,photoreal,epic realistic,dark shot,shadows,darkness,contrast,layered colors,vivid colors,contrast,offcial art,colorful,splash of color,movie perspective,very aesthetic,disheveled hair,perfect composition,moist skin,intricate details,moody,epic,photorealistic,color graded cinematic,atmospheric lighting,award winning photo,film grain,A shot with tension,Visual impact,giving the poster a dynamic and visually striking appearance,impactful picture,\",\n", " \"Voluptuous, attractive, 40 year old woman, thick glasses, big eyes, straight dark-brown collar-length hair, mascara, big nose\\ntight bellbottom trousers, tight blouse, huge tits, huge thighs, sharply-pleated flares\\nstanding in kitchen\\n\",\n", " \"from above, front view, 1girl, beautiful asian girl, harley quinn hair, long black hair with white highlights, caramel skin, standing, medium natural breasts, sexy pose, grey eyes, no makeup, looking away, toned legs, sexy ass, slim waist, long black tight dress, white dress, white choker, cleavage, breasts closeup, dark smoky solid background, Expressiveh\\n\\n\\n, sharp focus, stark lighting, medium-format camera, art by Helmut Newton, fashion photography, high-contrast, dramatic lighting, striking composition, confident and empowered woman, strong and assertive pose, fashion, sensuality, high end clothing, luxury accessories,dynamic pose, bdsm props,,\\n\\nigiex. \\n.. \\n. The background is minimalist, a gradient of shadowy blue tones, amplifying the ethereal and futuristic ambiance of the scene. The overall composition feels cinematic, blending high fashion with a science fiction aesthetic. . Exquisite cyberpunk. retroFuturism. . \\n Depth of field, low saturation,detail. mythp0rt Stylish, charismatic.\\n\\ndepth of field\\nmythp0rt,\\nnigiex.\",\n", " \"cosmic scene, cosmic coffee cup on a white table, inside cup is galaxies and stars of liquid marbles, liquid marble galaxy spilling out on table. astronaut holding mop, cleaning\",\n", " \"product photography of a cup of tea, zavy-sngkjd, masterpiece, award-winning, professional, highly detailed, atmospheric lighting, low key lighting, 100mm f/2.8 macro lens, fabric\",\n", " \"XUER guangying,gloves,holding,standing,outdoors,sky,helmet,robot,scenery,1other,science fiction,cable,ambiguous gender\\uff0cThis is a CGI-rendered image depicting a futuristic, alien landscape. The scene features two humanoid figures in space suits, standing on a rugged, rocky terrain. The figure on the left, wearing a dark, bulky space suit with a large, golden helmet, appears to be observing the scene. The figure on the right, closer to the foreground, is equipped with a more streamlined space suit in shades of yellow and black, with a helmet that reveals a face shield. This figure is holding a cylindrical device, possibly a communication or research tool. \\n\\nDominating the background is a massive, metallic spacecraft with intricate, angular designs and numerous cables and pipes extending from it. The craft's surface is a mix of white, grey, and black, with visible wear and tear, suggesting it has been in use for some time. The sky is a gradient of blue and grey, with wispy clouds, indicating a twilight or early morning setting. The foreground is characterized by a field of spiky, purple plants, adding texture and depth to the scene. The overall color palette is cool and muted, with a mix of blues, greys, and purples dominating the image. The scene exudes a sense of isolation and exploration, typical of science fiction narratives.\\n\\nhighly detailed,ultra-high resolution,32K UHD,sharp focus,best-quality,masterpiece,unconventional supreme masterpiece,masterful details,depth of field,blurry background,blurry foreground,motion blur,going really fast masterpiece,deep focus,\\nintricate details,soothing tones,high contrast,natural skin texture,soft light,sharp,masterpiece,best quality,ray tracing,sunlight,\\nultra-realistic,high dynamic range,photoreal,epic realistic,dark shot,shadows,This is a real high-definition photo,with professional photography and high resolution\\uff0c\\ncontrast,layered colors,vivid colors,contrast,offcial art,movie perspective,very aesthetic,disheveled hair,perfect composition,intricate details,moody,epic,photorealistic,\\ncolor graded cinematic,atmospheric lighting,award film grain,A shot with tension,Visual impact,giving the poster a dynamic and visually striking appearance,impactful picture,Depth of field, insanely interplay between lights and shadows,\"\n", "]\n", "```\n", "\n", "Cluster 4:\n", "```\n", "[\n", " \"b&w photo of 42 y.o man in black clothes, bald, face, half body, body, high detailed skin, skin pores, coastline, overcast weather, wind, waves, 8k uhd, dslr, soft lighting, high quality, film grain, Fujifilm XT3\",\n", " \"hyper-realism, full-high body view!! fashion 20yo luxury woman in motion, avant-garde fantasy, black, red, gold, wearing D&G transparent cyber punk style, expressive, ultra- sharp, white background,chiaroscuro light, vivid color, professional photo, intricate details\",\n", " \"zavy-hrglw, macro shot of a beautiful girl on a train, messy bun hair, neon yellow hair, silhouette, skinny, sexy, dimly lit, shadow, monochrome, b&w, instagram model, cinematic still, movie screencap, dynamic angle, 100mm f/2.8 macro lens, deep photo, depth of field, shadow,\",\n", " \" orange peel,, SFW, , best quality, , cinematic composition,extremely detailed,ultra-detailed,extremely detailed cg unity wallpaper,masterpiece,intricate details,cinemascope,high budget, low key lighting,foreshortening,back lit,vivid illumination,strong tones, during midnight,inside the Bellagio Conservatory & Botanical Gardens indoors in Las Vegas in front of a beautiful display of flowers, , ,, , SexScene3DRender, , , , ,\",\n", " \"painting of hybrid spider & cat & dragon & monkey, intercrossed animal, cat has eight legs, by zdzislaw beksinski, by lewis jones, by mattias adolfsson, cold hue's, warm tone gradient background, concept art, beautiful composition, digital painting, , \",\n", " \"Hyper realistic painting, earie, volumetric light, dealicate and breathtaking, picture of dancing cyborg girl goddess of light and purity on top of industrial ruins of frozen planet, robotic body cyborg, cyberpunk white robotic body suit, shining long golden hair, ground water reflections, neons shining on the night rain droplets, holy aura, anime artwork by grag rutkowski & mucha & wlop \",\n", " \"no humans, ultra detailed, black and white, outdoors, very wide shot, perspective, scene, scenary, top down view, stunning modern urban upscale environment, ultra realistic, concept art, elegant, extremely detailed, intricate, sharp focus, depth of field, professionally graded greyscale, volumetric fog, HDR, 32K, 16K, professional photograph with Nikon D90, 35mm lens, f/2.8 aperture, Lomography Lady Grey B&W 35mm ISO 400, movie promotional poster, vibrant details, in style of Ansel Adams BREAK beautiful outdoor \",\n", " \"girlfriend selfie, score_9, score_8_up, score_7_up, full-length character design, sexy but innocent lady, college girlfriend, undressing, eye contact, shy smile expression, bare perky breasts, small nipples, modern-style colour illustration, bold & powerful, lifelike art, cute college girl, blue & brown eyes, bikini, detailed setting, in a pool, in a california garden, exploring the surroundings, deep and colourful background, dark and moody lighting, green and blue and brown, clean girl aesthetic, blonde hair, e-girl makeup, clothes on floor, flirting, teasing, different poses, inset reaction shot, rich & seductive, freckles, reaction frame, erotic aesthetics\"\n", "]\n", "```\n", "\n", "Cluster 5:\n", "```\n", "[\n", " \"w00lyw0rld snail\",\n", " \"masterpiece, Phoenix,no human,Colorful feathers,fantasy, mythical,epic lighting, photo realism, high quality, highly detailed, masterpiece, epic,serenity\",\n", " \"Volcanic landscape, multiple active volcanoes erupting lava flows, lava rivers flowing through valleys, dark gray volcanic rock formations, orange-red hot lava flows, dark clouds of smoke and ash, patches of lush green vegetation, mountainous terrain, dramatic lighting, fiery red and orange color palette, atmospheric perspective, photorealistic style, high detail, 8k resolution, detailed lava textures\\n\\n\\n\\n\\n\\n\\naidmaMJ6.1,aidmafluxpro1.1\",\n", " \"masterpiece, score_9, score_8_up, score_7_up, EasyNegative_v2, \\nbest quality,ultra-detailed, high resolution, super detailed skin, beautiful detailed eyes, Brightly Lit, Highly Visible, 8k, sharp focus, perfect lighting, beautiful eyes, extremely detailed eyes, smooth edges, beautiful face, \\nseductive smile, smile, embarrassed, blush, \\n21 year old model, medium breasts, \\nleaning forward, foreshortening, \\nzzSuika, brown eyes, hair bow, horns, long hair, low-tied long hair, oni horns, orange hair, very long hair, horn ribbon, 1girl, long hair, white hair, long white dress, ballroom dress, white gloves, graveyard, purple theme, blue flames, spooky\",\n", " \"geometrical parental advisory amalgamation \",\n", " \"adorable cartoon style, a farmyard, friendly chicks playing in the farmyard, 3d render cartoon, made out of potato chips, , p1nkch1ps,\",\n", " \"assassinkahb style, a black and white photo of an assassin wearing a motorcycle jacket with japanese writing on the back, katana, solo, simple background, 1boy, white background, jacket, monochrome, upper body, greyscale, male focus, long hair in wind, weapon, clothes writing, skull, skeleton, japanese flag, atmospheric haze, Film grain, cinematic film still, shallow depth of field, highly detailed, high budget, cinemascope, moody, epic, OverallDetail, gorgeous, 2000s vintage RAW photo, photorealistic, candid camera, color graded cinematic, atmospheric lighting, skin pores, imperfections, natural, shallow dof,\",\n", " \"charmed, toucan, Within a shimmering sanctuary, a garden of dreams materializes, its iridescent blooms and shimmering petals a vision of pure fantasy, Reflecting off surfaces intricate details, abstract, whimsical, \"\n", "]\n", "```\n", "\n", "Cluster 6:\n", "```\n", "[\n", " \"masterpiece, best quality, 1 skull knight king, normal horse, sitting in horse, slimes_skullwm, made of skulls, \",\n", " \", score_9, score_8_up, score_7_up, slimes_skullwm, 1 tree, Socotra Dragon Tree, dry soil background, very huge tree, made of skulls\",\n", " \" slimes_skullwm, made of skulls,, a hinged bangle made out of\",\n", " \"score_9, score_8_up, score_7_up, Realistic piano made out of skull. stage, auditorium, very wide shot, orange spotlight, 3D\",\n", " \"score_9, score_8_up, score_7_up, slimes_skullwm, made of skulls, a woodland mansion from minecraft game made of skulls\",\n", " \"Masterpiece, score_9, score_8_up, score_7_up, ultra-detailed, slimes_skullwm, made of skulls, hedgehog\",\n", " \"score_9_up, score_8_up, score_7_up, medium shot, tyrannical king, wearing elaborate golden armour, sitting on a throne made of skulls, slimes_skullwm\",\n", " \"score_9, score_8_up, score_7_up, slimes_skullwm, hyper-realistic, photorealistic, ultra-detailed, a banana made of skulls\"\n", "]\n", "```\n", "\n", "Cluster 8:\n", "```\n", "[\n", " \"*Channel_42* *\\u4e2d*\\n*Hot springs* *thawing* *Autumn* *glow* *midnight* *serenity* *crystal* *bubbling* *red* *orange* *yellow* *green* *Altered* *Shard* *Spacetime*\",\n", " \"*Channel_42* *Broadcasting* *Woods* *Orc* *Stronghold* *Isometric* *Fantasy*\",\n", " \"*Channel_42* *\\u4e2d* *Station_42* *Broadcasting* \\n*Nightshade* *Midnight* *Ancient* *Hills* *Woods* *Sea* *Isolated* *Crimson*\\n*Shutting down* *Going Dark* *Offline*\",\n", " \"*Channel_42* *\\u4e2d* \\n *Postapocalyptic* *Ruins*\\n*Rough* *Rocky* *Ramparts* *Undulating* *Grass* *Classical* *Nostalgia* *Sideways* *Night* *Still* *light*\\n*Shutting Down*\\n*Going Dark*\\n*Offline*\",\n", " \"*Flow in the Name of Love*\\n*Channel_42* *\\u4e2d*\\n*Artificially_Atypically_Autistically*\\n*Icicle* *Ammolite* *Fossil Dun dun dun!!!*\\n*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*42*\\n*Shutting down*\\n*Going dark* \\n*Offline*\",\n", " \"*Channel_42* \\n *Jungle* *Pink* *Blue* *Violet* *Indigo* *Aloe* *Lotus* *Druidic* *Ruins* *Night* *Fireflies* *Glowing* *Cave* *Entrance* *Portal*\",\n", " \"*Flow in the Name of Love*\\n\\n*Channel_42* *\\u4e2d* *Broadcasting*\\n\\n*Artificially_Atypically_Autistically*\\n\\nhkstyle\\n\\n*Autumn* *Solstice* *Majesty* *Storm* *Irradiant* *Eventide* *Peak* *Silhouette* *Fantasy* *Dream* *Simple* *Twinkling* *Cosmic* *Crescent* *Orange* *Red* *Indigo* *Green* *Dark* *Illuminated*\\n\\n*Station_42*\\n\\n*Shutting down* *Going Dark* *Offline*\",\n", " \"*Channel_42* *Broadcast*\\n*TheFrozenNorth*\\n*Going dark*\\n*Shutting down*\\n*Offline*\\nmax details\"\n", "]\n", "```\n", "\n", "Cluster 10:\n", "```\n", "[\n", " \"masterpiece, best quality, newest, absurdres, \\nluo_tianyi,\\n1girl,shiny skin,blush,holding sword,chinese sword,dynamic angle,hanfu,chinese clothes,transparent clothes,tassel,chinese knot,bare shoulders,kanzashi,draped silk,gold trim,wind,bokeh,scattered leaves,flying splashes,waterfall,splashed water, BREAK looking at viewer,face focus,from side,sunrise, ray tracing,dynamic lighting,refined rendering,depth of field,ambience sense,detailed light,extremely light,beautiful detailed lighting,perfect lighting,best shadow,detailed shadow, \",\n", " \"score_9, score_8_up, score_7_up,source_anime, high res image,masterpiece,best quality,girl,cute face,clear skin,shiny hair,ultra detailed eyes,simple background, multicolored hair\",\n", " \"score_9,score_7_up,souce_anime,zPDXL2 BREAK\\nrip1ey,black hair,short hair,uniform,holster,hooded cloak,hood down,looking at viewer,black gloves, amputee, burn scar, \\ncowboy shot,dutch angle,\",\n", " \"score_4,score_5,score_6,score_7_up,score_8_up,score_9 masterpiece,best quality,ultra detailed,highres,HD,4k, \\n1 girl,solo,face focus,\\nFujiko.F.Fujio,,Eyelashes,\\nblue hair,glasses,red eyes,\\nCardigan, shirt,open clothes, long skirt,\\noutdoors,\\nsmile, blush,\",\n", " \"ailand,aerokinesis,1girl, solo,large breasts,breasts,blush,day,outdoors,canyon,white hair,red eyes, hair between eyes,looking at viewer, bangs, purple eyes, eyes visible through hair,very long hair, bangs,sidelocks,witch,witch hat,robe,skirt,,magic,holding_staff,wind,magic circle, gust,cyclone,action,dynamic pose, casting spell, twister,fluttering_clothes,anime_coloring,anime_screencap,studio_anime,\",\n", " \"score_9, score_8_up, score_7_up,source_anime, high res image,masterpiece,best quality,girl,cute face,clear skin,shiny hair,ultra detailed eyes,simple background, wolf cut hair,long hair\",\n", " \"score_4,score_5,score_6,score_7_up,score_8_up,score_9 masterpiece,best quality,ultra detailed,highres,HD,4k, \\n1girl, solo,solo focus,girl focus,face focus,\\nshort hair,brown hair,brown eyes,camisole,\\noutdoors,\\n, blowing kiss, pucked lips, heart, one eye closed, o3o, looking at viewer, hand up,Pouting lips,pointing fingertips towards viewer, blush,shy,smile,Half-closed eyes, squinting, staring,jitome,\",\n", " \"score_9,score_8_up,score_7_up,mecha,orange jacket,,orange peel,outdoors,orange theme,\"\n", "]\n", "```\n", "\n", "Cluster 11:\n", "```\n", "[\n", " \"score_9, score_8_up, score_7_up, source_anime, hina-dress, hina \\\\blue archive\\\\, medium breasts, puffy nipples, purple eyes, white hair,ponytail,horns,hair ribbon,low wings,halo,frilled dress,elbow gloves, open dress, trembling female, sweatdrop, embarrassed, nervous, open mouth, wavy mouth, swirly eyes, pov arms, adjusting another's hair\",\n", " \"Legit Free Buzz Inside...\",\n", " \"masterpiece, high quality, best quality, party, woman, queen, lingerie, mature, spoon, pregnant, fantasy, huge breasts, frilled dress,smiling, sleeves, frills, skirt, fat,looking to the side,talking, mini skirt, wet,eating semen,skimpy, cum stains, bukkake, cum on clothes, cum in mouth, smug eyes, cumflood, laughing, mouth opened, bowl filled with semen, gokkun, sitting around table, perspective, \",\n", " \"score_9,score_8_up,score_7_up, best quality, high quality, masterpiece, indoors, bedroom, source_anime,hetero, dark-skinned male, medium breasts, completely nude, \\nzPDXL2\\nmahirushiinauf, 1girl, long hair, yellow eyes, very long hair, blonde hair, on bed, \\n\\n lying, double penetration, 1girl, 2boys), anal, vaginal, on back, pussy, clitoris, anus, veiny penis, girl on top, from side, legs up, spread legs, cum in pussy, cum on anus, excessive, tongue out, ahegao,\",\n", " \"ultra detailed painting \\\\medium\\\\ by mzet, \\nvery awa, masterpiece, highres, absurdres, newest, year 2024, year 2023, studio shot, \\nsolo, albedo \\\\overlord\\\\, black elbow gloves, sitting backwards, crossed arms, arms on chair, lhead tilt, calm, closed eyes, smug, \\nsimple background, black theme, red theme, black background, eclipse, backlighting, \\n, , , , , ,\",\n", " \"league of legends, taur,action pose, full body, glowing claws, Extreme realistic depiction of a Wet Beauty Under Night Rain, in upper body shot, curvaceous glamorous body, hair over one eye, extremely wet, realistic wet skin texture, realistic wet clothing heavy rain, realistic fabric texture, high quality, masterpiece, taur, leopard, robot, cybernetics, power armor:, female, android, cyborg, shiny metal, dark blue metallic, muscular, black_nose, blue_eyes, black_sclera, machine landscape,bodysuit,mechanical tail, tail blade, cloudy sky, lightning, raining, rain, storm, masterpiece, best quality, newest, absurdres, highres, furry female, orange tiger girl, solo, biomechanical, tail, detailed, detailed paws, detailed hands, animal ears, digitigrade legs, short legs, no lineart, body silhouette, blue eyes glowing, detailed eyes, fighting stance, lowered, on one feet, showing claws, orange ice shard flying around, slit pupils, green eyes, orange ice shard flying around the body, white monochrome body, blue moon occupied half of background, above a layer of water, blue water, water relfections, detailed moon, detailed hands, full body view,cyberpunk, edgerunners,protogen, 1girl, female, woman, solo, anthro, furry, score_9,score_8_up,score_8_up,score_7_up,source_anime,source_furry,, ,Expressiveh, , ,g0thic, , ,csr style, ,\",\n", " \"city square, lamppost, 22 year old girl bound kneeling, elbows behind her back. skirt, shirt, boundhands, hands bound in front, cuffs collar,\",\n", " \"1girl, flower, neon lights, night, solo focus, purple hair, outdoors, looking at viewer, holding, breasts, hair bun, bouquet, cyberpunk, jacket, city, realistic, cleavage, pink flower, holding flower, lips, medium breasts, orange flower, blurry, nose, hair ornament, city lights, building, short hair, plant, yellow flower, closed mouth, hologram, standing, 1boy, , score_9, score_8_up, score_8_up, a_Comic_Illustration, rough lines with ink, drawn, sketch, crosshatching, comic style, graphic novel style, sharp lines, hires, zstyle, western comics \\\\style\\\\, hatching \\\\texture\\\\, zPDXL, Mr_Monster_Possitive\"\n", "]\n", "```\n", "\n", "Cluster 15:\n", "```\n", "[\n", " \"tr1ppystyle Kale\",\n", " \"\\n\\n\\n\",\n", " \"by Nicholas Roerich and Janek Sedlar and Alex Hirsch and Mikko Lagerstedt , surreal, retro, magical \",\n", " \"DSLR photo of \\u00c9\\u00b1\\u00c9\\u0090\\u00c5\\u00a7\\u00c9\\u00a6\\u00c7\\u009d\\u00ca\\u0089\\u00ca\\u0082\\u00c9\\u00b1\\u00c9\\u0090\\u00c5\\u0082\\u00c5\\u00a7\\u00c9\\u0090.in blade runner.photo.professional photography.high resolution.detailed photo.portrait.RAW.analog film.still film.50mm.f/16.uhd.hdr.4k \",\n", " \"\",\n", " \",\",\n", " \"capybara,clothes lift,selfie,\",\n", " \"Autumn. Otherworldly. bebop_style, ,\"\n", "]\n", "```\n", "\n", "Cluster 16:\n", "```\n", "[\n", " \"an atmospheric classical painting, dark night, full body view from a low angle, a towering undead knight holds a legendary war scythe, the blade describing a semicircle over his head, he is cloaked in an intricately decorated, ornate robe embroidered with occult symbols and arcane sigils, the imposing robe billows slightly around him, under the robe, he wears gleaming, ornate steel plate armor, including plate gloves and boots\\u00e2\\u0080\\u0094ancient artifacts from a long-gone era, decorative spiked chains hang loosely from his robe, his hollow eyes burn with an unnatural, cold fire beneath a shadowy hood, casting a faint glow on his skull, the blade of the scythe is engulfed in eldritch purple to blue flames that lick and twist through the air, casting surreal light and dancing shadows onto the cobblestone floor below, the knight's posture is rigid and powerful, the background consists of a castle's grand hall with towering gothic arches and arched windows with stained glass, the gothic architecture looms above, its vastness stretching upward into darkness, ominous shadows creep along the walls and arches, adding depth and tension to the scene, the primary light source comes from the eldritch flames engulfing his scythe and the knight's unnaturally burning eyes, the environment looks ancient and ominous, with crumbling stone walls overgrown with dark green moss and otherworldly bioluminescent leafy vines glowing faintly with purple light, a very low-angle perspective looks up at the skeletal knight, emphasizing his intimidating size and power, perfectly symmetrical architecture, intricate details, detailed skeletal head wearing a hood, detailed background, dark corners, style of fritz zuber-buhler, \",\n", " \"epic high fantasy scene, inspired by classic landscape artists with elements of romanticism, oil on canvas, low angle view emphasizing the dramatic scale, and a sense of awe in the presence of nature\\u00e2\\u0080\\u0099s grandeur, dappled sunlight, a secluded otherworldly valley in a mystical world, framed by steep, towering cliffs covered in lush, ancient flora, leading the eye toward a distant, grand spire that rises majestically on the horizon, in the foreground, a lone figure dressed in flowing, ornate black robes stands by a reflective pool, gazing toward the distant spire, the valley floor is scattered with moss-covered rocks and shallow pools that mirror the soft, filtered light from above, bioluminescent plants in shades of teal to cyan line the cliff edges, casting subtle blue glows that blend harmoniously with the warm, natural tones of the landscape, the distant spire is partly veiled in mist, evoking a mysterious, ancient quality, the hooded figure embodies the romantic spirit of man\\u00e2\\u0080\\u0099s smallness before nature, standing in quiet reverence, as if on a pilgrimage through a forgotten world, the figure wears a dark, flowing luxurious robe with a hood and holds a long ornate symmetrical staff with a large gleaming sapphire, adding an air of mystery, her stance is still and purposeful, suggesting a moment of introspection or awe as she takes in the sublime view before her, the format is a medium-long shot with a low angle view to enhance the towering cliffs and spire, creating a dramatic perspective that emphasizes the sheer scale of the landscape, with the viewer looking slightly upward, adding to the impression of vastness and grandeur, a soft, golden light suffuses the scene, reminiscent of early morning, casting gentle highlights across the foliage and creating a warm glow on the figure and rocks, the faint blue of the bioluminescent flora contrasts with the warm tones, adding a mystical quality to the scene, delicate mist drifts between the cliffs, enhancing the ethereal, dreamlike atmosphere, traditional oil on canvas, with rich, layered brushstrokes capturing the textures of the cliffs, foliage, and soft reflections on the water, subtle transitions between light and shadow evoke a sense of depth and mystery, high attention to detail in the plants and figure, while the distant elements are painted with a softer focus to draw attention toward the foreground, amanoer, style of charles-francois daubigny, \",\n", " \"highly detailed sci-fi cyberpunk digital artwork in the style of cknc, scifi, a teenage girl with cybernetic enhancements stands confidently in front of her towering mecha robot in a futuristic dystopian cyberpunk city, in the foreground, the girl\\u00e2\\u0080\\u0099s stance is firm and proud, framed by the glowing neon lights reflecting off her cybernetic enhancements and the mechanical armor of the colossal robot behind her, the cityscape is layered with holographic ads, gritty streets, and towering structures that fade into a misty, polluted skyline, distant, dimly glowing neon signs and billboards add depth, giving a sense of immense scale and urban density, the girl, outfitted with intricate cybernetic limbs, stands tall with a look of determination, her posture slightly angled to showcase both her profile and her powerful ally behind her, her attire is a mix of rugged and sleek, blending tactical urban gear with futuristic armor, her expression is fierce and focused, ready to face the dystopian world, medium shot from a low angle, emphasizing her stature and resilience while also capturing the massive scale of the robot and the sprawling city, the lighting is a blend of neon blues, purples, and pinks from holographic projections and distant signs, casting vibrant, reflected highlights on the mecha\\u00e2\\u0080\\u0099s metallic armor and the girl\\u00e2\\u0080\\u0099s enhancements, faint fog and dust filter through the light, adding to the dystopian atmosphere, subtle sparks and flickers from the mecha\\u00e2\\u0080\\u0099s exposed circuits hint at its recent use, high-resolution digital format, emphasizing fine detail in textures, metallic reflections, and the layering of light, intricate shading with contrast between warm streetlight hues and cool neon, evoking the dynamic, intense atmosphere of a cyberpunk cityscape, cyberkinetic enhancements, aidmamj6.1, \",\n", " \"magical atmosphere, nature focus, dim lighting, highly detailed masterwork painting portraying a beautiful witch, sfumato, golden ratio, a serene nighttime scene of a gorgeous nature witch casting a spectacular teal and gold-colored spell over a flower bed in her magical garden, she stands gracefully on a narrow cobblestone path in her lush herb garden, her captivating eyes focused on the spell, while her long, silky black hair flows softly in the night breeze, she wears a flowing, intricately detailed ornate robe in shades of deep purple and anthrazit grey, embroidered with floral motifs and a leather pouch filled with herbs at her side, her hands raised, tendrils of glowing green magic unfurling from her fingertips, illuminating the plants around her with a mystical light, the distant moon, partially obscured by dramatic clouds, is casting dappled moonlight across the land, the mood is peaceful and enchanting, the scene exuding a quiet sense of wonder, the herb garden is filled with fantastical plants, some bioluminescent, others with surreal shapes, a nearby plant features glowing blue flowers with petals resembling spiraled seashells, casting an ethereal light, another plant has long, twisting leaves that sway gently in response to the witch\\u00e2\\u0080\\u0099s magic, their edges faintly glowing purple, further back, a cluster of bulbous plants with translucent crimson roots pulse faintly, as if alive, contributing to the magical ambiance, the sky above is dramatic, with clouds drifting lazily, half obscuring the full moon, which bathes the garden in soft, silver dappled moonlight, casting gentle shadows and adding depth, the interplay of light and shadow enhances the dreamlike quality of the moment, the witch\\u00e2\\u0080\\u0099s face is serene, reflecting her deep connection to nature, the plants around her vibrant and alive in response to her magical spell, the perspective is intimate, as though the viewer is standing just a few feet away, watching the witch in her natural element, her garden framed by the deep shadows of the night, the foreground is rich in detail, with vibrant plants, the shimmering light of the magic, and the texture of earth and moss beneath her feet, the painting captures the warmth of the scene despite the cool night air, with subtle golden hues blending with the deep greens and purples of the garden, the background features shadowy trees and vines, fading into the dark, mysterious forest beyond, the overall color palette is rich and dark, dominated by deep greens, soft blues, purples, and silver moonlight accentuating the magic and lush garden, the oil painting technique adds depth and texture, with soft, layered brushstrokes giving a sense of movement and life to the garden, detailed perfect hands, detailed background, dark corners, hkmagic, style of gil elvgren, \",\n", " \"score_9,score_8_up,score_7_up,\\u7eaf\\u8272\\u80cc\\u666f\\uff0c\\u4e0a\\u534a\\u8eab\\uff0c\\u7535\\u5f71\\u53ca\\u7167\\u660e\\uff0c\\u4e2d\\u56fd\\u5973\\u5b69\\u5b50\\uff0c\\u77ed\\u53d1\\uff0c\\u5b8c\\u7f8e\\u7684\\u4eba\\u4f53\\u89e3\\u5256\\u5b66\\uff0c\",\n", " \"highly detailed dark fantasy-style painting of baba yaga, the witch, in front of her iconic hut on chicken legs, inspired by j.m.w. turner's shade and darkness - the evening of the deluge, eerie and foreboding atmosphere, romanticism, impressionism, surrealism, crepuscular rays, a mist-shrouded ancient primeval forest with otherworldly vegetation, thick with twisted, gnarled trees that loom menacingly over a swamp, tendrils of fog snake along the ground, dense underbrush cloaked in shadows, the landscape is wild, overgrown, and untamed, poisonous creeping vines and roots sprawled in every direction, and amidst the mist stands baba yaga\\u00e2\\u0080\\u0099s infamous hut perched on gigantic, grotesque chicken legs, the hut appears crooked and unstable, yet it towers above the swamp, its twisted wooden structure covered in decaying moss and old, weathered runes, skulls and ominous talismans hanging from the roof\\u00e2\\u0080\\u0099s edge, the foreground is filled with gnarled roots, glowing mushrooms, bioluminescent poisonous-looking swamp flowers, mossy rocks, and decaying leaves, in the background, ancient primeval trees stretch upwards into the mist, their branches clawing at the sky, the oppressive fog wraps everything in a suffocating grip, with baba yaga\\u00e2\\u0080\\u0099s hut standing out in stark contrast, glowing faintly from within, baba yaga stands in the foreground, an old hag, her silhouette hunched, her long grey hair blowing wildly in the night breeze, she is cloaked in tattered robes of dark, earthy tones, her bony hands clutching a twisted wooden staff decorated with bone charms and feathers, her piercing eyes glow with a malevolent light as she gazes at the viewer, her expression twisted with a knowing, sinister grin of dark amusement, her hut\\u00e2\\u0080\\u0099s massive chicken-like legs are mid-step, bent as it moves through the swamp, surrounded by swirling mist, the wooden structure ancient and weathered, the door slightly ajar as though beckoning to anyone foolish enough to approach, low-angle wide shot, the viewer looks up at baba yaga\\u00e2\\u0080\\u0099s imposing figure and the towering hut, emphasizing their size and power, the hut looms ominously in the background while baba yaga stands near the foreground, her hut\\u00e2\\u0080\\u0099s massive legs stepping through the swamp, casting ripples in the murky water below, the low angle creates a sense of looming danger and an overwhelming presence, the scene is dimly lit by the soft glow of the moon, with dappled moonlight filtering through clouds, faint beams of moonlight create dappled shadows across the swamp, the hut itself glows faintly from within, with eerie green and yellow light casting ominous reflections on the murky swamp waters, subtle backlighting from the moon gives baba yaga\\u00e2\\u0080\\u0099s figure an ethereal yet malevolent outline, faint glowing sparks of magic trail from baba yaga\\u00e2\\u0080\\u0099s staff, swirling with the mist, enhancing the dark, mystical feel, dark fantasy aesthetic with surreal elements, emphasizing mood and atmosphere, textures of decaying wood, damp earth, and creeping mist are prominent, creating a vivid sense of place and foreboding, fredfraistyle, \",\n", " \"classic painting, dappled lighting, 17th century setting, highly atmospheric masterwork painting, a rustic, dimly lit cabin with a stone hearth in the center, the scene illuminated by the flickering glow of a fire and soft beams of moonlight streaming through a small window on the side, casting shadows across the room, the floor is made of worn, rustic stone, adding to the ancient, timeworn feel of the setting, the foreground centers on a large iron cauldron hanging from a hook over a fire made of burning wooden logs surrounded by weathered granite stones, below it, the granite stone floor reflects the orange glow of the flames, shelves filled with mysterious ingredients line the walls in the background, casting an air of secrecy and old-world magic, dried herbs hang from the worn wooden beams above, wisps of steam rise from the cauldron, swirling lazily into the air, adding a misty, ethereal quality, while moonlight cuts through the scene with crepuscular rays, the subject is a gorgeous young witch, wearing a flowing black robe embroidered with silver, her full figure is visible, standing tall beside the cauldron, her robe sweeping down to the floor, stirring the bubbling liquid with a long, gnarled wooden staff, her attire is elegantly draped, the silver embroidery glinting faintly in the firelight, her face is serene and focused, her sharp eyes glinting in the low light as she watches her potion brew, steam curling around her, adding to the eerie atmosphere, her graceful posture contrasts with the dark, rustic surroundings, a full body shot from a low angle, looking slightly upward at the witch, emphasizing her tall, commanding presence within the room, the cauldron, the fire, and the granite stone floor all in view, the perspective highlights the space between her and the cauldron, creating a sense of depth and allowing the viewer to take in the full, eerie environment, dim, ambient lighting, with the fire beneath the cauldron serving as the main light source, casting a warm, flickering glow across the scene, the soft beams of moonlight entering through the small window add a cold, silvery contrast, highlighting her robe, the granite stone floor, and the surrounding ingredients in faint, ethereal light, subtle crepuscular rays cut through the steam, deepening the mysticism of the moment, oil on canvas, rich textures and soft, layered brushstrokes, capturing the glow of firelight and the cool highlights of moonbeams, a subdued color palette with deep blacks, dark browns, silvers, shades of green from the herbs, and shades of orange from the fire, delicate lighting effects enhance the magical atmosphere of the witch's quiet ritual, style of clyde caldwell, \",\n", " \"highly detailed oil painting of an otherworldly and hauntingly beautiful dense alien forest landscape featuring intricate otherworldly vegetation and nebula in the sky, the style blends gothic horror with ethereal, dreamlike fantasy and surrealism, a mind-bending, bewildering alien landscape with bizarre alien trees under a silver sky filled with strange, swirling, luminous clouds, the terrain is covered in dark, gnarled trees with twisted, contorted branches reaching toward the sky, in the distance, jagged mountains loom like the bones of an ancient creature, their peaks shrouded in mist and glowing vapors moving slowly across the land, in the foreground, strange, alien plants with blossoms resembling wide, unblinking eyes grow from the soil, some of the plants drip thick, viscous sap from serrated, blade-like leaves, while others glow with eerie blue light, a river of a sluggish black liquid winds its way through the landscape, shimmering with iridescent hues, in the background, a forest of towering gnarled trees with branches like monstrous claws looms, mysterious eldritch glowing vapors swirl around their roots, an eerie, alien world where every element feels as if it was waiting in ambush, the gnarled trees are covered in twisted vines and glowing moss, their bark forming grotesque gaping mouths, thick, serrated leaves drip glowing sap onto the ground, forming small pools of iridescent shimmering liquid, the landscape is shot in a wide-angle, that captures the scene in a panoramic view, the viewer is placed at the edge of the forest, soft, eerie lighting permeates the scene, glowing vapors swirl across the landscape, casting ethereal light and long shadows, the light from the strange plants and trees casts eerie reflections on the surface of the dark, iridescent river, faint glowing trails drift through the air, oil on canvas, with intricate detailing, the color palette blends deep purples, inky blacks, and shades of green with eerie, glowing highlights of blues, purples, and red, galaxy, nebula in the sky, \"\n", "]\n", "```\n", "\n", "Cluster 19:\n", "```\n", "[\n", " \"beretgirl, 1girl, long hair, beret, solo, purple hair, makeup,earrings lipstick,smile, green eyes, looking at viewer, score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, bed, nude,lying ,on stomach, crossed arms, naughty face, simple background, breasts, convenient censoring, ass, dark,bedroom,\",\n", " \"score_9, score_8_up, score_8, solo, solo, fluffy, feral, furry, mouse, female, vivid red eyes, grey fur, black ears, claws, fluffy cheeks, black dress, witch hat, cute, in the style of beatrix potter, flat color, backlit, high angle view, outdoors, magic circle, look at viewer, cruel expression, open mouth, talking, dynamic pose, hold magical staff,\",\n", " \"score_9,score_8_up,score_7_up, double anal,anal,group sex, 1girl,2boys,mmf threesome,female focus, gangbang,suspended congress, leg grab, standing, completely nude, cum in ass,ahegao, fucked silly\",\n", " \"score_9, score_8_up, score_7_up, 1boy, solo, tanaka, buzz cut, black hair, tank top, shorts, sportswear, gym, grin, sitting, lying, from front, front view, masculine, manly, from below,, anime screencap,\",\n", " \"score_9, score_8_up, score_7_up, score_6_up, source_anime, 2d, 1girl, 1boy, faceless male, facejerk, cum in face, tatsumaki one punch man, facial, fully nude, pout, head grab, male masturbation, shiny skin, cumshot in face, excessive cum, bedroom background, green eyes, \",\n", " \"score_9, score_8_up, score_7_up, score_6_up, source_anime, 1girl, long hair, \\nadjusting clothes, thong, huge ass,\\n\",\n", " \"score_9,score_8_up,score_7_up,\\nmonsterworld, masterpiece,high resolution,detail,\\n no humans, glowing, extra eyes, bike\\n, \",\n", " \"score_9, score_8_up, score_7_up, score_6_up, source_anime, \\n1girl, blonde hair, \\nupskirt, looking through legs, polka dot dress, cleft of venus, ass, breasts, nude,\"\n", "]\n", "```\n", "\n", "Cluster 20:\n", "```\n", "[\n", " \"score_9, score_8_up, score_8, score_9, dark goddess, perfect face, smooth skin, big devil horns, long hair, glowing eyes, cleavage, cinematic lighting, dark, seductive, perfect anatomy, focus, detailed, high definition, bedroom, Expressiveh, g0thicPXL, concept art, realistic, low-key, chiaroscuro\",\n", " \"safe_pos, Expressiveh, score_9, score_8_up, score_7_up, score_6_up 1 girl, 1 boy, mature woman, brown hair, blonde wavy hair, moaning, closed eyes, seductive face, naked, curvy, hourglass body, naturally sagging huge breast, perfect huge round ass, wide hips, thick thigh, pussy lips, pussy, wet pussy, long legs, stocking, high heel, wet body:1.2, sweat:1.2, BREAK \\n20yo, hard cock, \\nBREAK\\n in the living room, sex, alternate pose:1.5, pose variation:1.5, cock in pussy,\",\n", " \"score_9, score_8_up, score_7_up, score_6_up, 1girl, gorgeous girl, cute girl , kawai girl , cute face , nerd, round glasses, bangs , bob cut , medium firm breast , cleavage, puffy nipples, floral skirt, standing shy, boob focus, halfbody portrait, sexy body, sexy thighs,\",\n", " \"score_9,score_8_up,core_7_up,bleached,dark-skinned female, interracial, light-skinned male,dark skin, 1girl, tattoo, 1boy, breasts, penis, mirko, uncensored, sex, testicles, hetero, vaginal, tongue, sweat, navel, pussy, large breasts, rabbit ears, animal ears, girl on top, solo focus, long hair, tongue out, open mouth, white hair, spread legs, head out of frame, straddling, heart, saliva, teeth, day, interracial, cowgirl position, outdoors, pussy juice, sky, piercing\",\n", " \"score_9, score_8_up, score_7_up, realistic, 1girl, 1boy, blonde girl half asleep lying on pillow, close view, face focus, deepthroat, gagging suddenly awake, shocked, beautiful mouth, cock deep in mouth, face cum cumdrip from mouth, , pulling lips, pajamas, hair pull easynegative, cum covered face, excessive amount of cum, cum spurt,\",\n", " \"score_9,score_8_up,score_7_up,score_6_up, 1girl, lying on her side, in bed, medium breasts, spider Gwen, multicolored hair, ultra photorealistic, pink nipples, tiny waist, exposed breasts, close up, arched back, head on pillow, pov, topless, wet pussy, ripped Black leggings, wet spot, grool, messy hair, hair covering one eye,\",\n", " \"score_9, score_8_up, score_7_up, score_6_up, score_5_up, rich color, 1girl, solo, concept art, masterpiece, medium breasts, pose, sexy, parted lips, looking at viewer, laying on bed, looking up at ceiling, one knee up, head on pillow, arms up, view from above looking down, skimpy black nightgown, Expressiveh, spider gwen, tired, sweaty, at night, multicolored asymmetric shaved blond pink hair,\",\n", " \"score_9, score_8_up, score_7_up, score_6_up, gwzs, detailed drawing, 1girl, 1male out of frame, 20 year old, two-toned hair, black and red hair, perky breasts, cleavage, pretty face, petite, purple eyes, light freckles, oversized hoodie, black skirt, striped thigh-high socks, gothic boots, exhausted, looking up, saliva leaking from mouth, cock slumped over girl face, cock worship, tongue out, penis on face detailed background, outside, ally, graffiti, night, volumetric lighting, vivid colours, glowing, neon, full body shot from above\"\n", "]\n", "```\n", "\n", "Cluster 21:\n", "```\n", "[\n", " \"score_9, score_8_up, score_7_up, \\n1girl, tan, blonde hair, side ponytail, brown eyes, large breasts,\\n\\nschool uniform, blue shirt, long sleeves, sailor collar, bowtie, blue skirt, \\n\\nlying, on back, on bed, looking at viewer, clenched teeth, tears,\",\n", " \"score_9, score_8_up, score_7_up, score_6_up, source_anime, 1girl, solo, pmtgg, cyan hair, short hair, blunt bangs, green eyes, black and white dress, two-tone dress, grey bodysuit, bodysuit under clothes, grey pants, large breasts, looking at you, upper body, night sky, city, night, smile, portrait\",\n", " \"score_9, score_8_up, score_7_up, score_6_up, source_anime, 1girl, solo, cremia, red hair, long hair, pointy ears, big breasts, yellow t-shirt, jeans, looking at you, hand on hip, wink, peace sign, blue sky, city\",\n", " \"Rei-kun Style, \\nscore_9, score_8_up, score_7_up, rating_safe, 1boy, solo, male focus, mature male, orc, green skin, tusks, blue eyes, short hair, black hair, facial hair, beard, mustache, looking at viewer, armor, shoulder armor, breastplate, pauldrons, upper body, closed mouth, standing, outdoors, night, night sky, dark background\",\n", " \"general,highres, ultra-detailed,very aesthetic,best quality ,best hands, BREAK , 1990s \\\\style\\\\,anime coloring, \\nYuuki_Mizuho_OVA, 1girl,solo,long hair, brown hair, white hairband, red eyes, open mouth, blue ribbon,\\nschool uniform,red skirt, breasts, red blazer,\\noutdoor, day, sky, school,\\nsmile,looking at viewer, cowboy_Shot,\",\n", " \"Stable_Yogis_PDXL_Positives \\n score_9, score_8_up, score_7_up, score_6_up, 1girl, solo, breasts, looking at viewer, short hair, black hair, navel, closed mouth, sitting, medium breasts, nipples, collarbone, nude, water, dark-skinned female, completely nude, lips, muscular, night, arm support, abs, moon, tan, breasts apart, full moon, freckles, toned, tanlines, tomboy\",\n", " \"score_9, score_8_up, score_7_up, source_anime BREAK 1girl, solo, retro artstyle, pants, 1990s style, sidelocks, 1980s style, short hair with long locks, smile, open mouth, long sleeves, turtleneck, blue hair, blue eyes, window\",\n", " \"score_9, score_8_up, score_7_up, BREAK,\\n1girl, solo, BREAK,\\nsentaikageblue, shiny clothes, spandex, , BREAK,\\nsilver bodysuit, fishnet bodysuit, bodysuit under clothes, aqua dress, shoulder pads, aqua skirt, black belt, aqua gloves, no headwear, BREAK,\\nbrown hair, blunt bangs, long hair, brown eyes, medium breasts, thick thighs, BREAK,\\nfull-face blush, smug, grin, blush, looking at viewer, standing, waving, arm at side, BREAK,\\ndutch angle, outdoors, city, street, day, blue sky, cloudy sky\"\n", "]\n", "```\n", "\n", "Cluster 22:\n", "```\n", "[\n", " \"score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, \\n\\n,\\n\\nm0use_goldie, grey fur, green eyes, grey ears, mouse,\\n\\n,\\n\\nferal, cute, tail, blue dress, scarf, dark grey hooded cloak, arms crossed, looking at viewer, standing, open mouth, feet,\\n\\n,\\n\\noutdoors, forest, snow, wind, head focus, high angle view, from above, simple background,\",\n", " \"score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up,\\n\\nmouse, feral, female, solo, fluffy, fluffy cheeks, cute, \\n\\nfrilly shirt, blue skirt, grey fur, grey ears, green eyes, \\n\\nflat color, in the style of Beatrix Potter,\\n\\ncel-shading, backlit, cool color pallet, upper body portrait, head focus, side view, tired expression, annoyed expression, potions, tea, bottles, look down, working, looking down, mouse sitting behind the desk with her elbows on it, head in hand, \\n\\nindoors, candlelight, cottage, desk, workshop, dark, night\",\n", " \"score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, \\n \\nm0use_goldie, cute, grey fur, green eyes, expressive, dynamic pose,\\n \\nsolo, frilly shirt, feral, upper body shot, head focus, indoors, cheerful expression, behind desk, papers, paperwork, clutter, side view, holding potion bottles, shelves, bottles, potions, open mouth,\",\n", " \"score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, \\n \\norange fur, grey eyes, cat, \\n \\nsolo, red tunic, white sash, white hooded cape, blue amulet, cute, feral, look at viewer, open mouth, desert, tail, upper body shot, detailed background, arms folded\",\n", " \"a fantasy:cyberpunk landscape\",\n", " \"score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up,\\n\\nmouse, feral, female, solo, fluffy, fluffy cheeks, cute\\n\\ncel-shading, in the style of Beatrix Potter,\\n\\nsleeveless red dress, yellow fur, yellow ears, blue eyes, sitting, side view, grass, nature, close up, upper body shot, looking at viewer, nervous, embarrassed expression\",\n", " \"score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, \\n\\n,\\n\\nsolo, m0use_goldie, peeking out upper body, red dress,\\n\\n,\\n\\ncute, feral, nervous expression, look at viewer, open mouth,\\n\\n,\\n\\ndoorway, cottage, rustic, cozy, indoors, high angle view,\",\n", " \"score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up,\\n\\nmouse, feral, furry, female, solo, fluffy, fluffy cheeks, cute, \\n\\nred dress, yellow fur, yellow ears, blue eyes, \\n\\nflat color, in the style of Beatrix Potter, peeking out upper body\\n\\ncel-shading, backlit, warm color pallet, upper body portrait, head focus, close up, side view, look at viewer, detailed fur,\\n\\noutdoors, behind tree, head peeking out, blush,\"\n", "]\n", "```\n", "\n", "Cluster 23:\n", "```\n", "[\n", " \"score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up,beautiful , masterpiece, flowers, emerald path, forest, made of orange peel, female with intricate detailed orange armor,\",\n", " \", Jawa-PXL, jawa, 1other, faceless, solo, musclegut, thigh focus, town, crack of light, , sandals, , crossed legs, zoom layer, score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, source_cartoon\",\n", " \"score_9, score_8_up, score_7_up, score_6_up, score_5_up, bkomadori, solo, vaporeon, underwater,\",\n", " \"score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, source_cartoon, 1girl, zzAndroid18, blue eyes, blonde hair, short hair,\",\n", " \"score_9, score_8_up, score_7_up, score_6_up, score_5_up, f33f4l, 1girl, long hair, horror \\\\theme\\\\,\",\n", " \"score_9, score_8_up, score_7_up, score_6_up, score_5_up, pryw, 1girl, long hair, portrait, demon, sweater\",\n", " \"score_9, score_8_up, score_7_up, foxfolk, a witch\",\n", " \"score_9, score_8_up, score_7_up, score_6_up, score_5_up, paperlarva, 1girl, gothic,\"\n", "]\n", "```\n", "\n" ] } ], "source": [ "for k in sorted(by_cluster.keys()):\n", "\tif len(by_cluster[k]) < 64:\n", "\t\tcontinue\n", "\n", "\tprompts = random.sample(by_cluster[k], 8)\n", "\tprompts = [clean_prompt(p) for p in prompts]\n", "\n", "\tprint(f\"Cluster {k}:\\n```\")\n", "\tprint(json.dumps(prompts, indent=2))\n", "\tprint(\"```\")\n", "\tprint()" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [] } ], "metadata": { "kernelspec": { "display_name": "tmpenv5", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.11.9" } }, "nbformat": 4, "nbformat_minor": 2 }