Voxa Studio

Open-source text-to-speech starter inspired by jamiepine/voicebox.

This project gives you a deployable baseline with:

• FastAPI backend
• Studio-grade browser UI with dual generation modes
• Single + batch MP3 generation endpoints
• Voice search/filter (search, locale, gender, limit)
• Preset script templates
• Persistent render history (replay / copy link / delete / clear-all)
• Batch report panel (success/failure per line)
• Batch progress indicator and pass-rate meter
• Realtime waveform visualizer in output player
• Runtime stats panel
• Desktop packaging pipeline for macOS and Windows
• Docker support
• GitHub Actions CI

UI Preview

Demo Features

• GET /api/health health check
• GET /api/voices list voices with filtering
• GET /api/presets built-in script presets
• GET /api/history list generation history
• DELETE /api/history/{id} delete a history record and audio file
• DELETE /api/history clear all history records and generated files
• GET /api/stats aggregate counters for dashboard
• POST /api/speak synthesize speech and store history
• POST /api/speak/batch generate up to 20 lines in one request

Stack

• Python 3.11
• FastAPI + Uvicorn
• edge-tts
• Plain HTML/CSS/JS frontend

Quick Start

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements-dev.txt
uvicorn app.main:app --reload --host 0.0.0.0 --port 8000

Open: http://127.0.0.1:8000

API Example

curl -X POST http://127.0.0.1:8000/api/speak \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Hello from Voxa Studio",
    "voice": "en-US-AriaNeural",
    "rate": 0,
    "pitch": 0,
    "format": "mp3"
  }'

Response:

{
  "id": "...",
  "voice": "en-US-AriaNeural",
  "locale": "en-US",
  "rate": 0,
  "pitch": 0,
  "format": "mp3",
  "audio_url": "/audio/<id>.mp3",
  "created_at": "2026-03-02T07:00:00Z",
  "characters": 29,
  "text_preview": "Hello from Voxa Studio"
}

Batch example:

curl -X POST http://127.0.0.1:8000/api/speak/batch \
  -H "Content-Type: application/json" \
  -d '{
    "texts": ["First line", "Second line"],
    "voice": "en-US-AriaNeural",
    "rate": 0,
    "pitch": 0,
    "format": "mp3"
  }'

Run Tests

pytest -q

Docker

docker build -t voxa-studio .
docker run --rm -p 8000:8000 voxa-studio

Deploy

Or deploy with Render Blueprint:

render blueprint launch

Desktop Downloads

Desktop binaries are published as release assets:

• VoxaStudio-macOS-arm64.zip
• VoxaStudio-windows-x64.zip

When a GitHub release is published, .github/workflows/desktop-release.yml builds both targets and uploads the assets automatically.

Run desktop build locally:

pip install -r requirements-desktop.txt pyinstaller
python desktop/build.py

Project Layout

.
├── app/
├── desktop/
├── output/playwright/
├── render.yaml
├── web/
├── tests/
└── .github/workflows/

Safety Notes

• Do not use generated audio for impersonation, fraud, or deceptive use.
• Add user consent checks before deploying voice-clone features.
• If you add uploads or auth later, enforce rate limits and storage lifecycle controls.

License

MIT