Spaces:
Running
on
Zero
Running
on
Zero
import gradio as gr | |
import torch | |
import yt_dlp | |
import os | |
import subprocess | |
import json | |
from threading import Thread | |
from transformers import AutoTokenizer, AutoModelForCausalLM | |
import spaces | |
import moviepy.editor as mp | |
import time | |
import langdetect | |
import uuid | |
HF_TOKEN = os.environ.get("HF_TOKEN") | |
print("Starting the program...") | |
model_path = "Qwen/Qwen2.5-7B-Instruct" | |
print(f"Loading model {model_path}...") | |
tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True) | |
model = AutoModelForCausalLM.from_pretrained(model_path, torch_dtype=torch.float16, trust_remote_code=True).cuda() | |
model = model.eval() | |
print("Model successfully loaded.") | |
def generate_unique_filename(extension): | |
return f"{uuid.uuid4()}{extension}" | |
def cleanup_files(*files): | |
for file in files: | |
if file and os.path.exists(file): | |
os.remove(file) | |
print(f"Removed file: {file}") | |
def download_youtube_audio(url): | |
print(f"Downloading audio from YouTube: {url}") | |
output_path = generate_unique_filename(".wav") | |
ydl_opts = { | |
'format': 'bestaudio/best', | |
'postprocessors': [{ | |
'key': 'FFmpegExtractAudio', | |
'preferredcodec': 'wav', | |
}], | |
'outtmpl': output_path, | |
'keepvideo': True, | |
} | |
with yt_dlp.YoutubeDL(ydl_opts) as ydl: | |
ydl.download([url]) | |
# Check if the file was renamed to .wav.wav | |
if os.path.exists(output_path + ".wav"): | |
os.rename(output_path + ".wav", output_path) | |
if os.path.exists(output_path): | |
print(f"Audio download completed. File saved at: {output_path}") | |
print(f"File size: {os.path.getsize(output_path)} bytes") | |
else: | |
print(f"Error: File {output_path} not found after download.") | |
return output_path | |
def transcribe_audio(file_path): | |
print(f"Starting transcription of file: {file_path}") | |
temp_audio = None | |
if file_path.endswith(('.mp4', '.avi', '.mov', '.flv')): | |
print("Video file detected. Extracting audio...") | |
try: | |
video = mp.VideoFileClip(file_path) | |
temp_audio = generate_unique_filename(".wav") | |
video.audio.write_audiofile(temp_audio) | |
file_path = temp_audio | |
except Exception as e: | |
print(f"Error extracting audio from video: {e}") | |
raise | |
print(f"Does the file exist? {os.path.exists(file_path)}") | |
print(f"File size: {os.path.getsize(file_path) if os.path.exists(file_path) else 'N/A'} bytes") | |
output_file = generate_unique_filename(".json") | |
command = [ | |
"insanely-fast-whisper", | |
"--file-name", file_path, | |
"--device-id", "0", | |
"--model-name", "openai/whisper-large-v3", | |
"--task", "transcribe", | |
"--timestamp", "chunk", | |
"--transcript-path", output_file | |
] | |
print(f"Executing command: {' '.join(command)}") | |
try: | |
result = subprocess.run(command, check=True, capture_output=True, text=True) | |
print(f"Standard output: {result.stdout}") | |
print(f"Error output: {result.stderr}") | |
except subprocess.CalledProcessError as e: | |
print(f"Error running insanely-fast-whisper: {e}") | |
print(f"Standard output: {e.stdout}") | |
print(f"Error output: {e.stderr}") | |
raise | |
print(f"Reading transcription file: {output_file}") | |
try: | |
with open(output_file, "r") as f: | |
transcription = json.load(f) | |
except json.JSONDecodeError as e: | |
print(f"Error decoding JSON: {e}") | |
print(f"File content: {open(output_file, 'r').read()}") | |
raise | |
if "text" in transcription: | |
result = transcription["text"] | |
else: | |
result = " ".join([chunk["text"] for chunk in transcription.get("chunks", [])]) | |
print("Transcription completed.") | |
# Cleanup | |
cleanup_files(output_file) | |
if temp_audio: | |
cleanup_files(temp_audio) | |
return result | |
def generate_summary_stream(transcription): | |
print("Starting summary generation...") | |
print(f"Transcription length: {len(transcription)} characters") | |
detected_language = langdetect.detect(transcription) | |
prompt = f"""Summarize the following video transcription in 150-300 words. | |
The summary should be in the same language as the transcription, which is detected as {detected_language}. | |
Please ensure that the summary captures the main points and key ideas of the transcription: | |
{transcription[:300000]}...""" | |
response, history = model.chat(tokenizer, prompt, history=[]) | |
print(f"Final summary generated: {response[:100]}...") | |
print("Summary generation completed.") | |
return response | |
def process_youtube(url): | |
if not url: | |
print("YouTube URL not provided.") | |
return "Please enter a YouTube URL.", None | |
print(f"Processing YouTube URL: {url}") | |
audio_file = None | |
try: | |
audio_file = download_youtube_audio(url) | |
if not os.path.exists(audio_file): | |
raise FileNotFoundError(f"File {audio_file} does not exist after download.") | |
print(f"Audio file found: {audio_file}") | |
print("Starting transcription...") | |
transcription = transcribe_audio(audio_file) | |
print(f"Transcription completed. Length: {len(transcription)} characters") | |
return transcription, None | |
except Exception as e: | |
print(f"Error processing YouTube: {e}") | |
return f"Processing error: {str(e)}", None | |
finally: | |
if audio_file and os.path.exists(audio_file): | |
cleanup_files(audio_file) | |
print(f"Directory content after processing: {os.listdir('.')}") | |
def process_uploaded_video(video_path): | |
print(f"Processing uploaded video: {video_path}") | |
try: | |
print("Starting transcription...") | |
transcription = transcribe_audio(video_path) | |
print(f"Transcription completed. Length: {len(transcription)} characters") | |
return transcription, None | |
except Exception as e: | |
print(f"Error processing video: {e}") | |
return f"Processing error: {str(e)}", None | |
print("Setting up Gradio interface...") | |
with gr.Blocks(theme=gr.themes.Soft()) as demo: | |
gr.Markdown( | |
""" | |
# 🎥 Video Transcription and Smart Summary | |
Upload a video or provide a YouTube link to get a transcription and AI-generated summary. HF Zero GPU has a usage time limit. So if you want to run longer videos I recommend you clone the space. Remove @Spaces.gpu from the code and run it locally on your GPU! | |
""" | |
) | |
with gr.Tabs(): | |
with gr.TabItem("📤 Video Upload"): | |
video_input = gr.Video(label="Drag and drop or click to upload") | |
video_button = gr.Button("🚀 Process Video", variant="primary") | |
with gr.TabItem("🔗 YouTube Link"): | |
url_input = gr.Textbox(label="Paste YouTube URL here", placeholder="https://www.youtube.com/watch?v=...") | |
url_button = gr.Button("🚀 Process URL", variant="primary") | |
with gr.Row(): | |
with gr.Column(): | |
transcription_output = gr.Textbox(label="📝 Transcription", lines=10, show_copy_button=True) | |
with gr.Column(): | |
summary_output = gr.Textbox(label="📊 Summary", lines=10, show_copy_button=True) | |
summary_button = gr.Button("📝 Generate Summary", variant="secondary") | |
gr.Markdown( | |
""" | |
### How to use: | |
1. Upload a video or paste a YouTube link. | |
2. Click 'Process' to get the transcription. | |
3. Click 'Generate Summary' to get a summary of the content. | |
*Note: Processing may take a few minutes depending on the video length.* | |
""" | |
) | |
def process_video_and_update(video): | |
if video is None: | |
return "No video uploaded.", "Please upload a video." | |
print(f"Video received: {video}") | |
transcription, _ = process_uploaded_video(video) | |
print(f"Returned transcription: {transcription[:100] if transcription else 'No transcription generated'}...") | |
return transcription or "Transcription error", "" | |
video_button.click(process_video_and_update, inputs=[video_input], outputs=[transcription_output, summary_output]) | |
url_button.click(process_youtube, inputs=[url_input], outputs=[transcription_output, summary_output]) | |
summary_button.click(generate_summary_stream, inputs=[transcription_output], outputs=[summary_output]) | |
print("Launching Gradio interface...") | |
demo.launch() | |