ModelForce

One API for local and remote text-to-speech providers.

Switch between Piper, Kokoro, XTTS (and more) without changing your application.

Conceptually, ModelForce is for text-to-speech what LiteLLM is for language models.

✅ Local inference · ✅ Offline · ✅ Multiple providers · ✅ CLI · ✅ SDK · ✅ REST API · ✅ Web UI

Why?

Every TTS engine has its own API, installation, configuration, and voice management.

ModelForce provides one interface for all of them.

Without ModelForce	With ModelForce
Each provider has different API	One unified API
Separate auth for each	Single auth system
Separate configs	One config format
Vendor lock-in	Provider abstraction
Manual installation	One command install

30-Second Example

import { ModelForceClient } from "@modelforce/sdk";

const mf = new ModelForceClient();

await mf.synthesize({
    provider: "piper",
    text: "Hello world"
});

Change the provider. Your code didn't.

Works With

Provider	Type	Status
Piper	Local	✅ Full
Kokoro	Local	✅ Full
XTTS	Local	✅ Full

Getting Started

git clone https://github.com/PeepSick/ModelForce.git
cd ModelForce
pnpm install
pnpm build
pnpm --filter @modelforce/cli start -- doctor
pnpm --filter @modelforce/cli start -- pull piper
pnpm --filter @modelforce/cli start -- pull voice/piper/tr_TR-dfki-medium
pnpm --filter @modelforce/cli start -- voices
pnpm --filter @modelforce/cli start -- synthesize "Hello"
# Or with Kokoro (local ONNX inference):
pnpm --filter @modelforce/cli start -- pull kokoro
pnpm --filter @modelforce/cli start -- synthesize "Hello" --provider kokoro
# Or with XTTS (local Coqui TTS; requires: pip install TTS):
pnpm --filter @modelforce/cli start -- pull xtts
pnpm --filter @modelforce/cli start -- synthesize "Hello" --provider xtts

Architecture

Your App → SDK → Provider → Backend → Audio

Quick Commands

modelforce doctor                     # Health check
modelforce pull piper                 # Install Piper provider
modelforce pull kokoro                # Install Kokoro (local ONNX)
modelforce pull xtts                  # Install XTTS (Coqui TTS local)
modelforce pull voice/piper/tr_TR-dfki-medium  # Install voice
modelforce voices                     # List installed voices
modelforce status                     # Show providers/models/voices status
modelforce synthesize "Hello"         # Generate audio
modelforce quick "Hello"              # Quick synthesis
modelforce play "Hello"               # Synthesize and play

Why Not...

• LangChain - Focuses on LLM chains, not TTS runtime
• LiteLLM - Focuses on LLM routing, not TTS providers
• OpenAI SDK - Designed for OpenAI services; ModelForce focuses on provider-independent TTS runtimes

Packages

Package	Description
@modelforce/sdk	Simple API for text-to-speech
@modelforce/server	REST API server
@modelforce/web	React Web UI
@modelforce/cli	Command-line interface
@modelforce/core	Core types, health monitoring, event bus
@modelforce/runtime	Execution engine with parallel synthesis
@modelforce/character	Character profiles and voice presets
@modelforce/speech	Speech runtime (LeoVoice)
@modelforce/plugin	Plugin system with lifecycle management
@modelforce/registry	Provider and voice registry
@modelforce/installer	Provider installation management
@modelforce/provider-piper	Piper TTS provider
@modelforce/provider-kokoro	Kokoro TTS provider
@modelforce/provider-xtts	XTTS TTS provider

Known Issues

See Known Limitations.md

For Developers

pnpm test

Real integration tests run automatically when local models are present (Kokoro ONNX, XTTS model directory); otherwise they skip.

License

MPL 2.0