Skip to content

UseVoice/FrenchTTS

BranchesTags

Repository files navigation

FrenchTTS

Python 3.10+ Windows edge-tts faster-whisper Twitch/OBS AGPL-3.0

Realistic French TTS for Windows — no API key, no subscription, no compromise on voice quality.
Built for VoiceChat via VB-Cable or any virtual audio device.
Includes an optional STT pipeline (mic → faster-whisper → speech) for hands-free dictation.
Streaming-ready with a built-in Twitch / OBS mode — live overlay, channel-point TTS, and a local REST API.

Why FrenchTTS?

Finding a good French TTS in 2026 is surprisingly painful.

Most free solutions sound robotic, rely on outdated speech engines, or require you to paste text into a web interface and manually download the output. Paid alternatives (ElevenLabs, Azure, Google Cloud) do sound great — but they all require an API key, a credit card, and start billing once you exceed a free tier.

On the open-source side, offline models like Kokoro or XTTS exist, but they demand a GPU, several GB of model weights, and non-trivial setup just to get started.

FrenchTTS bridges that gap. It uses Microsoft Edge's neural TTS engine — the same one powering the Read Aloud feature built into Edge browser — through the edge-tts library. No account, no key, no install beyond pip. The voices are genuinely neural-quality, indistinguishable from a paid service for most use cases, and the latency is low enough for real-time roleplay.

The app wraps it in a clean dark-mode desktop UI with audio device routing, so you can pipe the output directly into voicechat (Discord, FiveM, TeamSpeak, ...) via VB-Cable without touching any config file or third-party tool.

Features

  • Neural voices — 8 French voices (France, Belgium, Canada, Switzerland) streamed from Microsoft Edge, no key required
  • Device selector — route audio to any output, auto-detects VB-Cable
  • Monitor output — optionally hear the TTS in your headphones simultaneously on a second device
  • Voice controls — adjustable speed, volume, and pitch
  • Input history — navigate previous texts with ↑ / ↓ (shell-style)
  • Replay — one-click or configurable hotkey (default F2) replays the last speech
  • Global hotkeys — Replay, Stop, and STT work even when the app is not focused
  • STT pipeline — mic → VAD → faster-whisper → TTS, hands-free dictation (disabled by default)
    • Energy-based VAD: starts recording automatically when you speak, stops on silence
    • Configurable keybind (default F1) shown on the button
    • Auto-restart mode: re-activates the mic after each TTS playback
    • Optional tray notification showing the transcribed text
  • System tray — closing the window hides to tray (balloon notification confirms); restores on double-click; quit via tray menu
  • Acrylic blur — Windows 10/11 native background blur with adjustable opacity
  • Resource controls (Performances) — CPU core throttle (affinity mask), Windows process priority (Normal / Below Normal / Idle), and RAM working-set soft cap; all live, all saved to config
  • Twitch / OBS mode (disabled by default, zero overhead when off)
    • OBS Browser Source overlay — real-time word-by-word karaoke display, synchronized with audio playback via WebSocket
    • Local REST API — POST /api/speak, /api/voice, /api/pitch, GET /api/status; callable from StreamElements, Streamlabs, curl, or any tool
    • Integrated Twitch bot — connects directly via IRC/PubSub; reacts to channel-point redemptions by title (TTS, Voix TTS, Pitch TTS, …)
    • Configurable appearance — overlay background (color, opacity, toggle), text color, all live-synced
    • Per-feature toggles — disable overlay, speak, voice, or pitch independently to save resources
  • Persistent config — all settings and history saved in %APPDATA%\UseVoice\FrenchTTS; atomic writes prevent corruption on crash
  • Auto-updater — checks GitHub Releases at launch, downloads FrenchTTSInstaller.exe, self-replaces silently; versioned by commit SHA (prod-XXXXXXX)
  • Installer — dark-themed CTk installer with progress steps; creates Desktop shortcut, Start Menu folder, and FrenchTTSUninstaller.exe
  • Uninstaller — removes app, config, STT models, shortcuts, and Start Menu entries
  • What's New dialog — shown once after each update with the release changelog
  • Buildable as .exe — 3-step PyInstaller build via build.bat

Voices

Display name Region Gender Voice ID
FR - Vivienne France Female fr-FR-VivienneMultilingualNeural
FR - Denise France Female fr-FR-DeniseNeural
FR - Eloise France Female fr-FR-EloiseNeural
FR-BE - Charline Belgium Female fr-BE-CharlineNeural
FR-CA - Sylvie Canada Female fr-CA-SylvieNeural
FR-CH - Ariane Switzerland Female fr-CH-ArianeNeural
FR - Remy France Male fr-FR-RemyMultilingualNeural
FR-CA - Antoine Canada Male fr-CA-AntoineNeural

All voices are neural quality, streamed in real time from Microsoft Edge TTS servers.

Requirements

  • Windows 10 / 11
  • Python 3.10+
  • Internet connection (TTS voices stream from Microsoft's servers; STT runs fully offline)

Quick start

Dev mode (no updater splash)
git clone https://github.com/UseVoice/FrenchTTS.git
cd FrenchTTS
pip install -r requirements.txt
python main.py

Or use the helper scripts at the project root:

Script What it does
setup.bat Install all pip dependencies
launch - dev.bat Run python main.py directly, no updater splash
launch - update.bat Run with --update flag to simulate an update
launch - installer.bat Test the installer or uninstaller interactively

Build as .exe

build.bat

Three-step build — each artifact is a prerequisite for the next:

Step Output Notes
1 dist\FrenchTTS.exe Main app; git SHA injected as build ID
2 dist\FrenchTTSUninstaller.exe Tiny stdlib-only uninstaller
3 installer\dist\FrenchTTSInstaller.exe Bundles both exes; this is the release asset

To publish a release, trigger the Build & Release workflow manually from the GitHub Actions UI. It builds all three artifacts, tags the release prod-<sha>, and attaches FrenchTTSInstaller.exe as the only release asset.

Write versions/<sha>.md beforehand if you want a custom What's New changelog (auto-generated otherwise from conventional commits).

Usage

Action How
Speak text Type → Enter or click Parler
Insert newline Shift + Enter
Stop playback Arrêter or configured hotkey (default F3)
Replay last audio Redire (F2) or configured hotkey (default F2)
Navigate history ↑ / ↓ in the text box
Open settings ⚙ Paramètres
Hide to tray Click the window's close button (×)
Quit Right-click the tray icon → Quitter
Start / stop STT 🎙 STT (F1) button or configured hotkey (default F1)

Global hotkeys (Replay, Stop, STT) fire even when the app is minimized or not focused.

STT — Speech-to-Text

STT is disabled by default. Enable it in ⚙ Paramètres → STT — Reconnaissance vocale → Activer.

STT uses faster-whisper (Whisper small model, CPU, int8) for offline French transcription.

Flow: press the STT button (or F1) → speak → silence is detected automatically → text is transcribed → TTS plays it back.

First use: the Whisper model (~460 MB) is downloaded once to %APPDATA%\UseVoice\FrenchTTS\stt_models and cached for all future sessions.

Settings:

Option Description
Activer Show/hide the STT button
Touche STT Keybind to toggle listening (default F1)
Redémarrage auto Automatically re-activates the mic after each TTS playback
Notif. texte Show a tray balloon with the transcribed text (title: "STT — Retranscription")
Microphone Input device for VAD/transcription

Twitch / OBS

Disabled by default. Enable in ⚙ Paramètres → Twitch / OBS → Activer.

When enabled, FrenchTTS starts a local HTTP server (default port 7681) that provides:

OBS Browser Source overlay

Add http://localhost:7681 as a Browser Source in OBS (recommended size: 1920 × 200 px).

Each time the TTS speaks, words appear one by one in sync with the audio, then fade out. The overlay uses a WebSocket connection and reconnects automatically.

Appearance is configurable in Paramètres → Twitch / OBS → Apparence de l'overlay:

Option Description
Fond sous texte Toggle background on/off
Opacité du fond Background transparency (0–100 %)
Couleur du fond Background colour (native colour picker or hex)
Couleur du texte Word colour

Changes are pushed live to all connected OBS sources — no refresh needed.

Local REST API

Method Route Body Description
POST /api/speak {"text":"…", "voice"?:"…", "pitch"?:N, "duration"?:N} Trigger TTS
POST /api/voice {"voice":"FR - Denise", "duration"?:30} Temporary voice change
POST /api/pitch {"pitch":-20, "duration"?:30} Temporary pitch change (Hz, −100→+100)
GET /api/status — {"speaking": bool, "voice":"…", "pitch":N}

voice and pitch overrides restore automatically after duration seconds (default: configurable in settings, default 30 s).

Integrated Twitch bot

Enable in Paramètres → Twitch / OBS → Bot Twitch intégré. Requires an OAuth token obtained from dev.twitch.tv (a guided setup is available directly in settings).

The bot subscribes to PubSub channel-point redemptions. Reward titles are matched case-insensitively:

Reward title Action
TTS / Lire TTS / Dire / Parler Reads the viewer's input
Voix TTS / Changer Voix / Voice TTS Changes voice temporarily
Pitch TTS / Changer Pitch Changes pitch temporarily (integer Hz)

Input text is capped at 300 characters per redemption.

Performances (resource management)

Available in ⚙ Paramètres → Performances.

Setting Description Default
Cœurs CPU Limits the process to N logical CPUs via SetProcessAffinityMask. Slider from 1 to the number of logical CPUs on your machine. 75 % of cores
Priorité CPU Windows process priority class — Normale, En dessous de la normale, or Basse (arrière-plan). Lower values reduce CPU scheduler share and may noticeably slow TTS generation. Normale
RAM max Soft working-set cap in MB via SetProcessWorkingSetSizeEx. Windows will page out memory above this limit when the system is under pressure. 4096 MB = Illimité (no cap). 1024 MB

All three settings are applied immediately on change and persisted to config.json.

VB-Cable setup

Example with Discord:

  1. Install VB-Audio Virtual Cable
  2. In FrenchTTS → ⚙ Paramètres → Sortie TTS → CABLE Input (VB-Audio Virtual Cable)
  3. In Discord → Settings → Voice → Input Device → CABLE Output (VB-Audio Virtual Cable)

To also hear the output in your own headphones, enable Casque in settings and select your headphone device.

Installation

Download FrenchTTSInstaller.exe from the latest release and run it.

The installer:

  • Extracts FrenchTTS.exe to %LOCALAPPDATA%\UseVoice\FrenchTTS\
  • Creates a Desktop shortcut
  • Creates a Start Menu folder (FrenchTTS) with shortcuts for the app and the uninstaller
  • Extracts FrenchTTSUninstaller.exe alongside the app

To uninstall, run Désinstaller FrenchTTS from the Start Menu (or FrenchTTSUninstaller.exe directly). It removes the app, all configuration, STT models, and shortcuts.

Data & file structure

%APPDATA%\UseVoice\FrenchTTS\
├── config.json            # voice, device, sliders, hotkeys, opacity, STT & performance settings
├── stt_models\            # faster-whisper model cache (downloaded on first STT use)
└── history\
    ├── last.mp3           # most recently generated audio
    └── lasts.log          # spoken text history (JSON array, max 100 entries)

%LOCALAPPDATA%\UseVoice\FrenchTTS\
├── FrenchTTS.exe          # installed application
└── FrenchTTSUninstaller.exe

FrenchTTS/
├── core/
│   ├── audio.py           # MP3 → PCM decoding, silence trimming
│   ├── constants.py       # paths, voices, settings defaults, formatters
│   ├── sounds.py          # STT audio feedback tones
│   └── version.py         # BUILD_ID — "dev" in source, SHA in release exe
├── installer/
│   ├── installer_main.py  # CTk installer + silent updater helper
│   ├── installer.spec     # PyInstaller spec (bundles app + uninstaller)
│   ├── uninstaller_main.py# Win32 stdlib uninstaller
│   └── uninstaller.spec   # PyInstaller spec (stdlib-only, small)
├── twitch/
│   ├── server.py          # aiohttp HTTP + WebSocket server (overlay, /api/* routes)
│   ├── manager.py         # TwitchManager lifecycle, feature flags, temp overrides
│   ├── bot.py             # twitchio IRC bot, PubSub channel-point handler
│   └── static/
│       ├── overlay.html   # OBS Browser Source — karaoke word-by-word display
│       └── callback.html  # OAuth redirect landing page — shows access_token to copy
├── ui/
│   ├── app.py             # FrenchTTSApp — main window + TTS pipeline
│   ├── settings.py        # SettingsWindow (collapsible sections, Twitch modals)
│   ├── updater.py         # UpdaterSplash, self-replacement logic
│   ├── utils.py           # window icons, acrylic blur, tray notifications, Win32 resource helpers
│   └── whats_new.py       # What's New dialog shown after updates
├── voice/
│   └── listener.py        # mic capture → VAD → faster-whisper STT pipeline
├── versions/
│   └── <sha>.md           # changelog shown in What's New after each update
├── img/
│   ├── icon.ico
│   ├── icon.png
│   └── logo.png
├── main.py
├── requirements.txt
├── build.bat
├── setup.bat
├── launch - dev.bat
├── launch - update.bat
└── launch - installer.bat

Dependencies

Package Role
edge-tts Neural TTS via Microsoft Edge servers
faster-whisper Offline French STT (Whisper model, CPU, int8)
customtkinter Modern dark-mode UI
sounddevice PCM playback with per-device routing
miniaudio In-memory MP3 decode (no ffmpeg needed)
numpy PCM buffer handling
keyboard System-wide hotkeys (works when app not focused)
pystray System tray icon and balloon notifications
Pillow Tray image and CTk internal rendering
aiohttp Local HTTP + WebSocket server for Twitch/OBS mode
twitchio Twitch IRC/PubSub bot for channel-point events

aiohttp and twitchio are only imported when Twitch mode is enabled. They add no startup overhead otherwise.

License

This project is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0).

© 2026 UseVoice

You are free to use, modify, and distribute this software, provided that:

  • The source code remains available.
  • Any modifications are released under the same AGPL-3.0 license.
  • If the software is used over a network (e.g., as a SaaS), the corresponding source code must be made publicly available.

For full details, see the LICENSE file or visit: https://www.gnu.org/licenses/agpl-3.0.html

About

Realistic French TTS for Windows — no API key, no subscription, no compromise on voice quality.

tts.usevoice.fr/

Resources

Readme

License

AGPL-3.0 license
Activity
Custom properties

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases 5

FrenchTTS prod-de6f34a Latest
Apr 19, 2026
+ 4 releases

Contributors

Languages