Spaces:
Sleeping
Sleeping
| import streamlit as st | |
| from downloader import download_video | |
| from extractor import extract_audio | |
| from detector import detect_accent | |
| import asyncio | |
| # Setup async event loop | |
| try: | |
| asyncio.get_running_loop() | |
| except RuntimeError: | |
| loop = asyncio.new_event_loop() | |
| asyncio.set_event_loop(loop) | |
| st.set_page_config(page_title="Accentometer", page_icon="🎙️") | |
| # Sidebar navigation | |
| st.sidebar.title("🎙️ Accentometer") | |
| page = st.sidebar.radio("", ["Home", "Info"]) | |
| # --------------------------- | |
| # Page 1: Home | |
| # --------------------------- | |
| if page == "Home": | |
| st.title("🎙️ Accentometer") | |
| # ✅ Example video links | |
| st.markdown("### Try with these example MP4 links:") | |
| st.text("Australian: https://docs.google.com/uc?export=download&id=12dKoO-jgWgjor_aQpovbtwOulMQBEEps") | |
| st.text("British: https://docs.google.com/uc?export=download&id=10aiY_0dnsWXqhxeNL54n5LYe6l4XquR8") | |
| st.text("American: https://docs.google.com/uc?export=download&id=1k1wfSxmQ-ZbYCKNZrb0aF8xfM6JWEfKQ") | |
| st.markdown( | |
| """ | |
| 📌 **Instructions**: Copy one of the above MP4 URLs and paste it into the field below, then click **Analyze** to detect the accent, and get a transcript. | |
| """ | |
| ) | |
| st.markdown(""" --- """) | |
| st.write("Paste a direct public MP4 video URL to **classify the English accent, get transcript**.") | |
| video_url = st.text_input("Direct Video URL (MP4 only):") | |
| if st.button("Analyze"): | |
| if not video_url: | |
| st.error("Please provide a valid MP4 video URL.") | |
| else: | |
| try: | |
| with st.spinner("Downloading video..."): | |
| video_file_path = download_video(video_url) | |
| with st.spinner("Extracting audio..."): | |
| audio_file = extract_audio(video_file_path) | |
| with st.spinner("Transcribing & detecting accent..."): | |
| accent, confidence, transcript = detect_accent(audio_file) | |
| st.success(f"Accent: **{accent}**") | |
| st.metric(label="Confidence", value=f"{confidence}%") | |
| st.text_area("Transcript", transcript, height=150, disabled=True) | |
| except Exception as e: | |
| st.error(f"Error: {str(e)}") | |
| # --------------------------- | |
| # Page 2: Info | |
| # --------------------------- | |
| elif page == "Info": | |
| st.title("ℹ️ About Accentometer") | |
| st.markdown(""" | |
| ### 🧠 What is Accentometer? | |
| **Accentometer** is a simple **demo** application for English accent classification. | |
| It doesn't use any advanced machine learning or deep learning models beyond basic heuristics. | |
| --- | |
| ### ⚙️ How It Works | |
| 1. **Video Downloader** | |
| Accepts only **direct MP4 links**. | |
| (Loom URLs aren’t supported in this free demo version — integration would require a paid Loom API or other backend setup.) | |
| 2. **Audio Extraction** | |
| Uses the `moviepy` library to extract the audio from the video. | |
| 3. **Speech-to-Text** | |
| Applies the `openai/whisper-base` model from Hugging Face to transcribe the audio into text. | |
| 4. **Accent Detection** | |
| Runs simple **hand-crafted heuristic rules** over the transcribed text to classify the accent into: | |
| - 🇺🇸 American | |
| - 🇬🇧 British | |
| - 🇦🇺 Australian | |
| - 🇮🇳 Indian | |
| - ❓ Unknown | |
| --- | |
| ### 🚀 Possible Improvements | |
| If you're interested in expanding this demo into a more robust application: | |
| - **Support More Input Options**: Upload local MP4 files, YouTube links, Loom links (via API), Additional video formats, etc. | |
| - **Improve Accent Detection**: Train a custom model using a labeled accent dataset, or integrate a model like `accent-id-commonaccent_xlsr-en-english`, or other approaches. | |
| --- | |
| """) | |