update readme

README.md

@@ -1,5 +1,5 @@
 ---
-title: Whisper Vs Sensevoice Small
+title: Whisper Vs SenseVoice Small
 emoji: ⚡
 colorFrom: gray
 colorTo: purple
@@ -8,28 +8,66 @@ sdk_version: 5.31.0
 app_file: app.py
 pinned: false
 license: mit
-short_description: Compare OpenAI Whisper against Sensevoice Small Resultssssss
+short_description: Compare OpenAI Whisper against FunAudioLLM SenseVoice
 ---
 
 # Whisper vs. FunASR SenseVoice Comparison
 
-This Space lets you compare OpenAI Whisper variants against FunAudioLLM’s SenseVoice models for automatic speech recognition (ASR), all via a simple Gradio 5 UI.
+This Space lets you compare OpenAI Whisper variants against FunAudioLLM’s SenseVoice models for automatic speech recognition (ASR), featuring:
 
-## 🚀 Demo
+- Support for multiple Whisper and SenseVoice models.
+- Language selection for each ASR engine.
+- **Speaker diarization enabled by default** to distinguish speakers in the transcription.
+- **Simplified Chinese to Traditional Chinese conversion enabled by default**.
+- Device selection for Whisper ASR (GPU or CPU).
+- Toggle punctuation output for SenseVoice.
+- Upload audio or record from microphone.
+- Side-by-side transcripts and diarized transcripts display.
 
-1. **Select Whisper model** from the dropdown.
-2. **Select SenseVoice model** from the dropdown.
-3. (Optional) **Toggle punctuation** for SenseVoice.
-4. **Upload** an audio file (wav, mp3, etc.) or **record** with your microphone.
-5. Click **Transcribe** to run both ASRs side-by-side.
+## 🚀 How to Use
+
+1. Select a **Whisper model** and its language.
+2. Choose device for Whisper: GPU or CPU.
+3. Select a **SenseVoice model** and its language.
+4. Optionally toggle punctuation for SenseVoice.
+5. Upload an audio file or record audio via microphone.
+6. Both ASRs run side-by-side with speaker diarization applied by default.
+7. Transcripts show both the plain text and speaker-labeled diarized text.
+8. Transcriptions convert Simplified Chinese to Traditional Chinese automatically.
 
 ## 📁 Files
 
 - **app.py**  
-  Main Gradio application. Sets up two HF-ASR pipelines and displays their outputs.
+  The Gradio app code implementing ASR and diarization pipelines with UI.
 
 - **requirements.txt**  
-  Python dependencies: Gradio, Transformers, Torch, Torchaudio, Accelerate, ffmpeg-python.
+  Lists Python dependencies including Gradio, Transformers, Torch, pyannote.audio, pydub, opencc-python-reimplemented, and others.
+
+- **README.md**  
+  This documentation file.
+
+## ⚠️ Notes
+
+- You must set your Hugging Face API token in the Space Secrets as `HF_TOKEN` for diarization to work properly (pyannote models are gated).
+- The app falls back to an older diarization pipeline version if the newer gated version is not accessible.
+- Simplified to Traditional Chinese conversion uses `opencc-python-reimplemented`.
+
+
+## 🛠️ Dependencies
+
+- Python 3.8+
+- PyTorch
+- Transformers
+- Gradio >= 5.0
+- pyannote.audio (for diarization)
+- pydub (audio processing)
+- opencc-python-reimplemented (Chinese script conversion)
+
+## License
+
+MIT License
+
+---
+
+If you find issues or want to suggest improvements, please open an issue or a PR.
 
-- **readme.md**  
-  This documentation.