changes

.ipynb_checkpoints/Untitled-checkpoint.ipynb

@@ -0,0 +1,33 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "564964b0",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import string\n",
+    "\n",
+    "original_string = \"This is a \\t test string with \\n whitespace.\"\n",
+    "\n",
+    "# Create a generator that filters out whitespace and count removed characters\n",
+    "filtered_chars = (char for char in original_string if char not in string.whitespace)\n",
+    "no_whitespace_string = \"\".join(filtered_chars)\n",
+    "removed_chars_count = sum(1 for char in original_string if char in string.whitespace)\n",
+    "\n",
+    "print(f\"String without whitespace: '{no_whitespace_string}'\")\n",
+    "print(f\"Number of removed characters: {removed_chars_count}\")\n"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3.12",
+   "language": "python",
+   "name": "python3"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

Untitled.ipynb

@@ -0,0 +1,54 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "id": "933bf3ba",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "String without whitespace: 'Thisisateststringwithwhitespace.'\n",
+      "Number of removed characters: 10\n"
+     ]
+    }
+   ],
+   "source": [
+    "import string\n",
+    "\n",
+    "original_string = \"This is a \\t test string with \\n whitespace.\"\n",
+    "\n",
+    "# Create a generator that filters out whitespace and count removed characters\n",
+    "filtered_chars = (char for char in original_string if char not in string.whitespace)\n",
+    "no_whitespace_string = \"\".join(filtered_chars)\n",
+    "removed_chars_count = sum(1 for char in original_string if char in string.whitespace)\n",
+    "\n",
+    "print(f\"String without whitespace: '{no_whitespace_string}'\")\n",
+    "print(f\"Number of removed characters: {removed_chars_count}\")\n"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3.12",
+   "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": "3.12.2"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

app.py

@@ -11,10 +11,11 @@ css = """
 
 with gr.Blocks(theme="base", css=css) as demo:
     gr.Markdown("<center><h1> 🔊 Transcription <span class='cursive-text'>Delight</span> </h1></center>")
-    gr.Markdown("### Step 1: Generate Raw Transcript")
     with gr.Row():
         with gr.Column():
-            source = gr.Radio(label="Source type", choices=[("Audio", "audio"), ("Video", "video"), ("YouTube URL", "youtube")], value="audio")
+            with gr.Row():
+                source = gr.Radio(label="Source type", choices=[ ("YouTube URL", "youtube"), ("Audio", "audio"), ("Video", "video")], value="youtube")
+                cleanup_options = gr.CheckboxGroup(label="Cleanup Transcript with LLM", choices=["Remove typos", "Separate into paragraphs"])
             @gr.render(inputs=source)
             def show_source(s):
                 if s == "audio":
@@ -31,46 +32,47 @@ with gr.Blocks(theme="base", css=css) as demo:
                         [source, source_component],
                         [download_audio],
                         show_progress="minimal"
+                    ).success(
+                        lambda : gr.Accordion(open=True),
+                        None,
+                        raw_accordion
                     ).then(
                         transcribe.transcribe,
                         [download_audio],
                         [preliminary_transcript],
-                    ).then(
-                        lambda : [gr.Button(interactive=True), gr.CheckboxGroup(interactive=True)],
+                        show_progress="minimal"
+                    ).success(
+                        lambda : (gr.Accordion(open=False), gr.Accordion(open=True)),
                         None,
-                        [clean_btn, cleanup_options]
+                        [raw_accordion, final_accordion]
+                    ).then(
+                        clean.clean_transcript,
+                        [download_audio, cleanup_options, llm_prompt, preliminary_transcript],
+                        [final_transcript, download_md],
+                        show_progress="minimal"
                     )
 
         with gr.Column():
             with gr.Row():
-                transcribe_btn = gr.Button("Transcribe audio 📜", variant="primary")
-                download_audio = gr.DownloadButton("Download .mp3 File 📥", interactive=False)
-            preliminary_transcript = gr.Textbox(info="Raw transcript", lines=10, max_lines=10, show_copy_button=True, show_label=False, interactive=False)
+                transcribe_btn = gr.Button("Transcribe 📜", variant="primary")
+            with gr.Accordion("Raw transcript", open=False) as raw_accordion:
+                preliminary_transcript = gr.Markdown("*Raw transcript will appear here*", show_label=False, height=400)
+            with gr.Accordion("Final transcript", open=False) as final_accordion:
+                final_transcript = gr.Markdown("*Final transcript will appear here*", height=400)
 
     source.change(utils.transcribe_button, source, transcribe_btn)
 
-    gr.Markdown("### Step 2: Clean with an LLM")
-    with gr.Row():
-        with gr.Column():
-            cleanup_options = gr.CheckboxGroup(label="Cleanup Transcript with LLM", choices=["Remove typos", "Separate into paragraphs"])
-            llm_prompt = gr.Textbox(label="LLM Prompt", visible=False, lines=3)
-            cleanup_options.change(
-                utils.generate_prompt,
-                cleanup_options,
-                llm_prompt
-            )
-            with gr.Row():
-                clean_btn = gr.Button("Clean transcript ✨", variant="primary")            
-                download_md = gr.DownloadButton("Download .md 📥", interactive=False)
-
-        with gr.Column():
-            final_transcript = gr.Markdown("*Final transcript will appear here*", height=400)
-
-    clean_btn.click(
-        clean.clean_transcript,
-        [download_audio, cleanup_options, llm_prompt, preliminary_transcript],
-        [final_transcript, download_md],
-        show_progress="minimal"
-    )
+    with gr.Accordion("⚙️ Settings and Files", open=False) as settings_accordion:
+        with gr.Row():
+            with gr.Column():
+                llm_prompt = gr.Textbox(label="LLM Prompt", visible=False, lines=3)
+                cleanup_options.change(
+                    utils.generate_prompt,
+                    cleanup_options,
+                    [llm_prompt, settings_accordion]
+                )
+                with gr.Row():
+                    download_audio = gr.DownloadButton("Download .mp3 File 📥", interactive=False)
+                    download_md = gr.DownloadButton("Download .md 📥", interactive=False)
     
 demo.launch()

clean.py

@@ -1,7 +1,6 @@
 from huggingface_hub import InferenceClient
 from pathlib import Path
 import gradio as gr
-import os
 
 MODEL_NAME = "meta-llama/Meta-Llama-3-70b-Instruct"
 
@@ -20,10 +19,10 @@ def clean_transcript(audio_file, options, prompt, transcript: str):
             messages = [
                 {"role": "user", "content": prompt + "\n" + chunk}
             ]
-            client = InferenceClient(model=MODEL_NAME, token=os.getenv("HF_TOKEN"))
+            client = InferenceClient(model=MODEL_NAME)
             for c in client.chat_completion(messages, max_tokens=1000, stream=True):
                 token = c.choices[0].delta.content
-                text += token
+                text += token or ""
                 yield text, None
         
     # write text to md file

utils.py

@@ -42,10 +42,7 @@ def convert_video_to_audio(input_file):
         gr.Error(f"An error occurred: {e}")
 
 def transcribe_button(source):
-    if source == "audio":
-        return gr.Button("Transcribe audio 📜")
-    else:
-        return gr.Button("Transcribe video 📜")
+    return gr.Button("Transcribe 📜")
 
 def generate_audio(source, source_file):
     if source == "audio":
@@ -61,11 +58,11 @@ def generate_audio(source, source_file):
 def generate_prompt(cleanup):
     prompt = "The following is a raw transcript from an automatic transcription system. "
     if not cleanup:
-        return gr.Textbox(visible=False)
+        return gr.Textbox(visible=False), gr.Accordion(open=False)
     elif "Remove typos" in cleanup:
         prompt += "Fix the minor typos (e.g. misspellings, homophones) in the transcript so that the transcript reads more logically. "
     if "Separate into paragraphs" in cleanup:
         prompt += "Separate the transcript into paragraphs to make it more readable. "
     prompt += "Don't add any extra words in your response, like 'Here is the corrected transcript:' just return the final transcript."
-    return gr.Textbox(visible=True, value=prompt)
+    return gr.Textbox(visible=True, value=prompt), gr.Accordion(open=True)