Windows SAPI Bridge for ElevenLabs — Use ElevenLabs TTS in Calibre & Any SAPI App

[xTaraRose]

Buy me a coffee · GitHub Repo · Sign up for ElevenLabs

Windows SAPI Bridge for ElevenLabs
Seamless ElevenLabs TTS — everywhere Windows speaks.

— Why I built this —

I was tired of trying to find different implementations to integrate ElevenLabs into different apps and always coming up short with nothing straightforward and simple to use. Especially with using Calibre a lot — there was realistically nothing out there.

So I present to you the Windows SAPI Bridge for ElevenLabs.

It works across the board with anything that uses the Windows SAPI, allowing seamless use of the ElevenLabs TTS engine and resulting in a much better listening experience — no per-app setup, no hacks, no workarounds. Install once, use everywhere.

— What it does —

Windows TTS is built on the SAPI5 (Speech API) standard. Every voice on your system — including Microsoft's built-in ones — is just a COM object registered in the Windows registry. Any application that uses SAPI for text-to-speech will automatically see and be able to use any registered voice.

This project:

1. Registers a lightweight Python COM server as a SAPI5 TTS engine
2. Queries your ElevenLabs account and creates one SAPI voice token per voice
3. Each voice appears in Windows' voice list as ElevenLabs - [name] — indistinguishable from a built-in voice
4. When an app speaks, the engine streams raw PCM audio from ElevenLabs directly to SAPI

Works with any SAPI5-compatible application, including:

• 📚 E-book readers (e.g. Calibre's Read Aloud)
• ♿ Screen readers and accessibility tools (e.g. NVDA, Windows Narrator)
• 🗣️ Dedicated TTS players (e.g. Balabolka, TTSReader)
• 🏠 Home automation / custom scripts using the Windows Speech SDK
• 🎮 Games and applications with built-in narration
• Any other software with a Windows TTS voice dropdown

— Requirements —

• Windows 10 / 11 (64-bit)
• Python 3.8+ (64-bit) — python.org
• ElevenLabs account (free tier works) — Sign up here
• Internet connection (audio is streamed in real time)

— Installation —

Step 1 — Install Python dependencies
Open a normal (non-admin) Command Prompt:

pip install comtypes requests

Step 2 — Add your ElevenLabs API key
Open config.json and replace YOUR_API_KEY_HERE with your ElevenLabs API key. Keep this file in the same folder as the scripts.

Step 3 — Run the installer as Administrator
Right-click Command Prompt → Run as administrator, then:

cd C:path	owindows-sapi-bridge-elevenlabs
python install_voices.py

This queries your ElevenLabs account and registers every voice in Windows automatically.

Step 4 — Select a voice in your app
Open any SAPI-compatible app, go to its TTS/voice settings, and all your ElevenLabs voices appear as ElevenLabs - [name]. Select one and it just works.

Note: The first sentence takes 1–2 seconds longer while Windows spins up the engine. Subsequent speech streams smoothly.

— Speed Control —

Set a baseline speed in config.json ("speed": 1.0, range 0.5–2.0). If your app has its own rate slider, the bridge honours it too and stacks it on top of your baseline — or disable that and use only your configured speed. Config changes take effect immediately, no reinstall needed.

— Uninstalling —

python uninstall_voices.py

Run as Administrator. Cleanly removes all SAPI registrations from Windows.

— Links —

• GitHub: https://github.com/xTaraRose/Windows...for-ElevenLabs
• ElevenLabs: Sign up / learn more (affiliate link)

Made with frustration and determination by xTaraRose

Buy me a coffee if this saved you some time!