Sliders / README.md
Rileypiper's picture
Upload 4 files
38b47f1 verified
|
Raw History Blame Contribute Delete
2.55 kB
---
title: Voice Target Slider
emoji: πŸŽ™οΈ
colorFrom: purple
colorTo: pink
sdk: gradio
sdk_version: "5.47.2"
python_version: "3.10"
app_file: app.py
pinned: false
---
# Voice Target Slider
A voice analysis tool for **gender-affirming voice therapy**. Upload a reference speaker's voice ("target") to see where it lands on five acoustic dimensions, then record your own voice to compare.
## Five Dimensions
| Slider | What it measures | Acoustic proxy | 0 end | 100 end |
|--------|-----------------|----------------|-------|---------|
| **Pitch** | Fundamental frequency | Median F0 (Hz) | Low | High |
| **Weight** | Vocal fold mass / breathiness | H1βˆ’H2 (dB) | Heavy | Light |
| **Resonance** | Vocal tract brightness | F2 βˆ’ F1 (Hz) | Dark | Bright |
| **Prosody** | Melodic variation | F0 SD (semitones) | Monotone | Bouncy |
| **Volume** | Loudness | RMS (dBFS), inverted | Loud | Soft |
## Workflow
1. **(Optional)** Set a personal baseline in Settings to map sliders relative to your own voice
2. **Target tab** β€” Upload or record a reference voice β†’ click **Analyze Target**
3. **Your Voice tab** β€” Read the Rainbow Passage (or another prompt) β†’ click **Analyze Recording**
4. Compare your sliders to the target. The **Delta** row shows the difference.
5. Use **Live mode** for continuous real-time feedback while practicing
## Mapping Modes
- **Global** β€” Broad default distributions for slider mapping
- **Anchored (literature)** β€” Uses wider speaking-range anchors from voice research literature
- **Baseline (personal)** β€” Maps slider positions relative to your own baseline recording
## Run Locally
```bash
pip install -r requirements.txt
python app.py
```
## Deploy to Hugging Face Spaces
1. Create a new Space β†’ select **Gradio** SDK
2. Upload `app.py`, `requirements.txt`, and this `README.md`
3. The Space starts automatically
## Exports
The app writes session data inside the container runtime:
- `exports/profile.json` β€” Latest analysis snapshot
- `exports/history.csv` β€” Full session history
- `exports/baseline.json` β€” Baseline mapping (if created)
## Notes & Limitations
- Metrics are sensitive to microphone quality, distance, room noise, and speaking task
- H1βˆ’H2 and formant estimates can be unstable for some recordings; the app includes robustness checks but is not clinical instrumentation
- For best results, speak continuously for 10+ seconds and use a quiet environment
- The Rainbow Passage is included as a standard reading prompt used in clinical voice assessment