Spaces:
Sleeping
Sleeping
| import os | |
| import json | |
| import torch | |
| import whisper | |
| import streamlit as st | |
| from tempfile import NamedTemporaryFile | |
| def transcribe_audio(audio_file, model): | |
| """ | |
| Transcribe a single audio file using OpenAI's Whisper model locally. | |
| """ | |
| result = model.transcribe(audio_file) | |
| return result["text"].strip() | |
| def main(): | |
| st.title("Audio Transcription with Whisper") | |
| # File uploader | |
| uploaded_file = st.file_uploader("Choose an audio file", type=["wav", "mp3", "m4a", "flac", "aac"]) | |
| if uploaded_file is not None: | |
| # Display audio file details | |
| file_details = {"Filename": uploaded_file.name, "FileSize": uploaded_file.size} | |
| st.write(file_details) | |
| # Play audio | |
| st.audio(uploaded_file, format='audio/wav') | |
| if st.button('Transcribe Audio'): | |
| with st.spinner('Transcribing audio using Whisper large model...'): | |
| # Check if CUDA is available | |
| device = "cuda" if torch.cuda.is_available() else "cpu" | |
| st.info(f"Using device: {device}") | |
| # Load the Whisper model | |
| model = whisper.load_model("large", device=device) | |
| # Save uploaded file temporarily and transcribe | |
| with NamedTemporaryFile(delete=False, suffix=os.path.splitext(uploaded_file.name)[1]) as tmp_file: | |
| tmp_file.write(uploaded_file.getvalue()) | |
| tmp_file_path = tmp_file.name | |
| transcription = transcribe_audio(tmp_file_path, model) | |
| # Remove temporary file | |
| os.unlink(tmp_file_path) | |
| # Display transcription | |
| st.subheader("Transcription:") | |
| st.write(transcription) | |
| # Save transcription to JSON | |
| output_json = 'transcription.json' | |
| with open(output_json, 'w', encoding='utf-8') as f: | |
| json.dump({uploaded_file.name: transcription}, f, ensure_ascii=False, indent=4) | |
| st.success(f"Transcription saved to {output_json}") | |
| if __name__ == "__main__": | |
| main() |