import os import google.generativeai as genai from moviepy.video.io.VideoFileClip import VideoFileClip import tempfile import logging import gradio as gr # Suppress moviepy logs logging.getLogger("moviepy").setLevel(logging.ERROR) # Configure Gemini API genai.configure(api_key=os.environ["GEMINI_API_KEY"]) # Create the Gemini model generation_config = { "temperature": 0.7, # Lower temperature for more accurate results "top_p": 0.9, "top_k": 40, "max_output_tokens": 8192, "response_mime_type": "text/plain", } model = genai.GenerativeModel( model_name="gemini-2.0-pro-exp-02-05", generation_config=generation_config, ) # List of all supported languages SUPPORTED_LANGUAGES = [ "Auto Detect", "English", "Chinese", "German", "Spanish", "Russian", "Korean", "French", "Japanese", "Portuguese", "Turkish", "Polish", "Catalan", "Dutch", "Arabic", "Swedish", "Italian", "Indonesian", "Hindi", "Finnish", "Vietnamese", "Hebrew", "Ukrainian", "Greek", "Malay", "Czech", "Romanian", "Danish", "Hungarian", "Tamil", "Norwegian", "Thai", "Urdu", "Croatian", "Bulgarian", "Lithuanian", "Latin", "Maori", "Malayalam", "Welsh", "Slovak", "Telugu", "Persian", "Latvian", "Bengali", "Serbian", "Azerbaijani", "Slovenian", "Kannada", "Estonian", "Macedonian", "Breton", "Basque", "Icelandic", "Armenian", "Nepali", "Mongolian", "Bosnian", "Kazakh", "Albanian", "Swahili", "Galician", "Marathi", "Punjabi", "Sinhala", "Khmer", "Shona", "Yoruba", "Somali", "Afrikaans", "Occitan", "Georgian", "Belarusian", "Tajik", "Sindhi", "Gujarati", "Amharic", "Yiddish", "Lao", "Uzbek", "Faroese", "Haitian Creole", "Pashto", "Turkmen", "Nynorsk", "Maltese", "Sanskrit", "Luxembourgish", "Burmese", "Tibetan", "Tagalog", "Malagasy", "Assamese", "Tatar", "Hawaiian", "Lingala", "Hausa", "Bashkir", "Javanese", "Sundanese" ] def extract_audio_from_video(video_file): """Extract audio from a video file and save it as a WAV file.""" video = VideoFileClip(video_file) audio_file = os.path.join(tempfile.gettempdir(), "extracted_audio.wav") video.audio.write_audiofile(audio_file, fps=16000, logger=None) # Suppress logs return audio_file def transcribe_audio_with_gemini(audio_file): """Transcribe audio using Gemini with a magic prompt for accurate timestamps.""" with open(audio_file, "rb") as f: audio_data = f.read() # Create proper audio blob audio_blob = { 'mime_type': 'audio/wav', 'data': audio_data } # Magic prompt for transcription with timestamps prompt = """ You are a professional transcriber. Transcribe this audio accurately and verbatim in the original language. Include timestamps for each sentence in the following format: [HH:MM:SS] Sentence 1 [HH:MM:SS] Sentence 2 don't change any in format ... Ensure the timestamps are accurate and correspond to the start of each sentence. Respond only with the transcription and timestamps. Do not add explanations or extra text. """ # Transcribe audio convo = model.start_chat() convo.send_message(prompt) response = convo.send_message(audio_blob) return response.text.strip() def generate_subtitles(transcription): """Generate SRT subtitles from transcription with timestamps.""" lines = transcription.split("\n") srt_subtitles = "" for i, line in enumerate(lines, start=1): if not line.strip(): continue # Extract timestamp and text if line.startswith("["): timestamp = line.split("]")[0] + "]" text = line.split("]")[1].strip() else: timestamp = "[00:00:00]" text = line.strip() # Convert timestamp to SRT format start_time = timestamp[1:-1] # Remove brackets start_seconds = time_to_seconds(start_time) end_seconds = start_seconds + 5 # Placeholder: 5 seconds per line end_time = seconds_to_time(end_seconds) srt_subtitles += f"{i}\n{start_time},000 --> {end_time},000\n{text}\n\n" return srt_subtitles def time_to_seconds(time_str): """Convert HH:MM:SS to seconds.""" hh, mm, ss = map(int, time_str.split(":")) return hh * 3600 + mm * 60 + ss def seconds_to_time(seconds): """Convert seconds to HH:MM:SS.""" hh = seconds // 3600 mm = (seconds % 3600) // 60 ss = seconds % 60 return f"{hh:02}:{mm:02}:{ss:02}" def translate_srt(srt_text, target_language): """Translate an SRT file while preserving timestamps using a magic prompt.""" # Magic prompt for translation prompt = f""" Translate the following SRT subtitles into {target_language}. Preserve the SRT format (timestamps and structure). Translate only the text after the timestamp. Do not add explanations or extra text. Ensure the translation is accurate and culturally appropriate. Here is the SRT file: {srt_text} """ response = model.generate_content(prompt) return response.text def process_video(video_file, language="Auto Detect", translate_to=None): """Process a video file to generate and translate subtitles.""" # Extract audio from the video audio_file = extract_audio_from_video(video_file) # Transcribe audio using Gemini transcription = transcribe_audio_with_gemini(audio_file) # Generate subtitles subtitles = generate_subtitles(transcription) # Save original subtitles to an SRT file original_srt_file = os.path.join(tempfile.gettempdir(), "original_subtitles.srt") with open(original_srt_file, "w", encoding="utf-8") as f: f.write(subtitles) # Translate subtitles if a target language is provided translated_srt_file = None if translate_to and translate_to != "None": translated_subtitles = translate_srt(subtitles, translate_to) translated_srt_file = os.path.join(tempfile.gettempdir(), "translated_subtitles.srt") with open(translated_srt_file, "w", encoding="utf-8") as f: f.write(translated_subtitles) # Clean up extracted audio file os.remove(audio_file) return original_srt_file, translated_srt_file, "Detected Language: Auto" # Define the Gradio interface with gr.Blocks(title="AutoSubGen - AI Video Subtitle Generator") as demo: # Header with gr.Column(): gr.Markdown("# 🎥 AutoSubGen") gr.Markdown("### AI-Powered Video Subtitle Generator") gr.Markdown("Automatically generate and translate subtitles for your videos in **SRT format**. Supports **100+ languages** and **auto-detection**.") # Main content with gr.Tab("Generate Subtitles"): gr.Markdown("### Upload a video file to generate subtitles.") with gr.Row(): video_input = gr.Video(label="Upload Video File", scale=2) language_dropdown = gr.Dropdown( choices=SUPPORTED_LANGUAGES, label="Select Language", value="Auto Detect", scale=1 ) translate_to_dropdown = gr.Dropdown( choices=["None"] + SUPPORTED_LANGUAGES[1:], # Exclude "Auto Detect" label="Translate To", value="None", scale=1 ) generate_button = gr.Button("Generate Subtitles", variant="primary") with gr.Row(): original_subtitle_output = gr.File(label="Download Original Subtitles (SRT)") translated_subtitle_output = gr.File(label="Download Translated Subtitles (SRT)") detected_language_output = gr.Textbox(label="Detected Language") # Link button to function generate_button.click( process_video, inputs=[video_input, language_dropdown, translate_to_dropdown], outputs=[original_subtitle_output, translated_subtitle_output, detected_language_output] ) # Launch the Gradio interface with a public link demo.launch(share=True)