Spaces:
Runtime error
Runtime error
diegopacheco
commited on
Commit
•
42bc318
1
Parent(s):
f0011e3
app v1
Browse files- app.py +99 -0
- comics.png +0 -0
- comics_0.png +0 -0
- e1_comics_0.png +0 -0
- e2_comics_0.png +0 -0
- e3_comics_0.png +0 -0
- install-deps.sh +3 -0
- requirements.txt +18 -0
- run.sh +3 -0
app.py
ADDED
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from diffusers import DiffusionPipeline
|
2 |
+
from torchvision.transforms.functional import to_tensor
|
3 |
+
import torch
|
4 |
+
import matplotlib.pyplot as plt
|
5 |
+
from torchvision.transforms.functional import to_pil_image
|
6 |
+
import gradio as gr
|
7 |
+
from torchvision.utils import save_image
|
8 |
+
from PIL import Image
|
9 |
+
import os
|
10 |
+
from gtts import gTTS
|
11 |
+
import torch
|
12 |
+
import gradio as gr
|
13 |
+
from torchvision.transforms import Compose, Resize, CenterCrop, ToTensor, Normalize
|
14 |
+
from transformers import pipeline, GPT2LMHeadModel, GPT2Tokenizer
|
15 |
+
|
16 |
+
example_1 = "ninja turtles fighting against a mosquito, in the sea"
|
17 |
+
example_2 = "warrior fighting zombies with a sword, in the forest"
|
18 |
+
example_3 = "western cowboy fighting against a dragon, in the desert"
|
19 |
+
|
20 |
+
def load_image(image_path):
|
21 |
+
images = []
|
22 |
+
image = Image.open(image_path)
|
23 |
+
image_tensor = to_tensor(image)
|
24 |
+
image_tensor = image_tensor / image_tensor.max()
|
25 |
+
images.append(image_tensor)
|
26 |
+
return to_pil_image(images[0])
|
27 |
+
|
28 |
+
def text_to_comics(text):
|
29 |
+
if text == example_1:
|
30 |
+
return load_image("e1_comics_0.png")
|
31 |
+
if text == example_2:
|
32 |
+
return load_image("e2_comics_0.png")
|
33 |
+
if text == example_3:
|
34 |
+
return load_image("e3_comics_0.png")
|
35 |
+
|
36 |
+
pipeline = DiffusionPipeline.from_pretrained("ogkalu/Comic-Diffusion")
|
37 |
+
output = pipeline(text, prompt_len=70, num_images=1, return_tensors=True)
|
38 |
+
|
39 |
+
images = []
|
40 |
+
for i in range(1):
|
41 |
+
image = output.images[i]
|
42 |
+
image_tensor = to_tensor(image)
|
43 |
+
image_tensor = image_tensor / image_tensor.max()
|
44 |
+
save_image(image_tensor, f"comics_{i}.png")
|
45 |
+
images.append(to_pil_image(image_tensor))
|
46 |
+
return images[0]
|
47 |
+
|
48 |
+
#text_to_comics(example_1)
|
49 |
+
#text_to_comics(example_2)
|
50 |
+
#text_to_comics(example_3)
|
51 |
+
|
52 |
+
def generate_story(description):
|
53 |
+
model = GPT2LMHeadModel.from_pretrained("gpt2")
|
54 |
+
tokenizer = GPT2Tokenizer.from_pretrained("gpt2")
|
55 |
+
inputs = tokenizer.encode(description + " a thriller/action story.", return_tensors='pt')
|
56 |
+
outputs = model.generate(input_ids=inputs,
|
57 |
+
max_length=200,
|
58 |
+
num_return_sequences=1,
|
59 |
+
temperature=0.7,
|
60 |
+
no_repeat_ngram_size=2)
|
61 |
+
story = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
62 |
+
return story
|
63 |
+
|
64 |
+
def convert_to_audio(text):
|
65 |
+
tts = gTTS(text)
|
66 |
+
audio_file_path = "audio.mp3"
|
67 |
+
tts.save(audio_file_path)
|
68 |
+
return audio_file_path
|
69 |
+
|
70 |
+
def audio_to_text(audio_file_path):
|
71 |
+
pipe = pipeline("automatic-speech-recognition", "openai/whisper-large-v2")
|
72 |
+
result = pipe("audio.mp3")
|
73 |
+
print(result)
|
74 |
+
return result['text']
|
75 |
+
|
76 |
+
def sentiment_analysis(text):
|
77 |
+
sentiment_analyzer = pipeline("sentiment-analysis")
|
78 |
+
result = sentiment_analyzer(text)
|
79 |
+
print(result)
|
80 |
+
return result
|
81 |
+
|
82 |
+
def app(text):
|
83 |
+
comics = text_to_comics(text)
|
84 |
+
story = generate_story(text)
|
85 |
+
audio_file = convert_to_audio(story)
|
86 |
+
transcribed_text = audio_to_text(audio_file)
|
87 |
+
sentiment = sentiment_analysis(transcribed_text)
|
88 |
+
return comics, audio_file,transcribed_text, sentiment
|
89 |
+
|
90 |
+
ui = gr.Interface(fn=app,
|
91 |
+
inputs="text",
|
92 |
+
outputs=["image", "audio", "text", "text"],
|
93 |
+
title="GenAI Multi-model LLM comics: Type some text get comics!",
|
94 |
+
description="This model generates comics based on the text(max 70 chars) you provide." + \
|
95 |
+
"<BR/>It does not work on mobile(timeout issue) click on examples if dont want to wait. " + \
|
96 |
+
"<BR/>It may take ~10-20min to generate the comics.",
|
97 |
+
examples=[(example_1),(example_2),(example_3)],
|
98 |
+
)
|
99 |
+
ui.launch()
|
comics.png
ADDED
comics_0.png
ADDED
e1_comics_0.png
ADDED
e2_comics_0.png
ADDED
e3_comics_0.png
ADDED
install-deps.sh
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
#!/bin/bash
|
2 |
+
|
3 |
+
/bin/pip install -r requirements.txt
|
requirements.txt
ADDED
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
numpy
|
2 |
+
transformers
|
3 |
+
sentence-transformers
|
4 |
+
seaborn
|
5 |
+
torch
|
6 |
+
torchvision
|
7 |
+
matplotlib
|
8 |
+
pandas
|
9 |
+
scikit-learn
|
10 |
+
nltk
|
11 |
+
gensim
|
12 |
+
tensorflow
|
13 |
+
keras
|
14 |
+
opencv-python
|
15 |
+
fastapi
|
16 |
+
uvicorn
|
17 |
+
gradio
|
18 |
+
|
run.sh
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
#!/bin/bash
|
2 |
+
|
3 |
+
/bin/python src/main.py
|