lexigram-multimedia-tts

Text-to-speech generation for the Lexigram Framework — local and API-based backends (local-http, elevenlabs, openai, chatterbox, kokoro, f5-tts, piper).

Overview

lexigram-multimedia-tts synthesizes speech from text. The default backend calls a local HTTP reference server (http://localhost:5002) so the package works out of the box with no API keys; hosted backends (ElevenLabs, OpenAI) and in-process reference servers (Chatterbox, Kokoro, F5-TTS, Piper) are selectable via config.

Full documentation: docs.lexigram.dev

Install

uv add lexigram-multimedia-tts
# Optional extras
uv add "lexigram-multimedia-tts[elevenlabs]"        # ElevenLabs API
uv add "lexigram-multimedia-tts[openai]"            # OpenAI TTS API
uv add "lexigram-multimedia-tts[chatterbox-server]" # local Chatterbox server (torch)
uv add "lexigram-multimedia-tts[kokoro-server]"     # local Kokoro server
uv add "lexigram-multimedia-tts[f5-tts-server]"     # local F5-TTS server (torch)
uv add "lexigram-multimedia-tts[piper-server]"      # local Piper server

Quick Start

from lexigram import Application
from lexigram.di.module import Module, module
from lexigram.multimedia.tts import AudioTTSModule
from lexigram.contracts.multimedia import TTSProvider, TTSRequest


@module(imports=[AudioTTSModule.configure()])
class AppModule(Module):
    pass


async def main() -> None:
    async with Application.boot(modules=[AppModule]) as app:
        tts = await app.container.resolve(TTSProvider)
        result = await tts.generate(
            TTSRequest(text="Hello from Lexigram", voice="alloy")
        )
        if result.is_ok():
            asset = result.unwrap()  # MediaAsset — audio bytes or URI


if __name__ == "__main__":
    import asyncio

    asyncio.run(main())

Configuration

Zero-config usage: Call AudioTTSModule.configure() with no arguments to use the local-http backend at http://localhost:5002.

Option 1 — YAML file

# application.yaml
multimedia:
  tts:
    backend: "elevenlabs"
    elevenlabs_voice_id: "21m00Tcm4TlvDq8ikWAM"

Option 2 — Profiles + Environment Variables

export LEX_PROFILE=production
export LEX_MULTIMEDIA__TTS__BACKEND=openai
export LEX_MULTIMEDIA__TTS__OPENAI_VOICE=echo

Option 3 — Python

from lexigram.multimedia.tts import AudioTTSModule
from lexigram.multimedia.tts.config import TTSConfig

AudioTTSModule.configure(
    config=TTSConfig(backend="elevenlabs", elevenlabs_voice_id="21m00Tcm4TlvDq8ikWAM")
)

Config reference

Module Factory Methods

Key Features

• Seven backends — local-http, elevenlabs, openai, chatterbox, kokoro, f5-tts, piper
• Reference servers — lexigram-tts-*-serve console scripts run each local model server
• Secret-managed API keys — provider keys resolved by name through the secrets backend
• Result-based — generate() -> Result[MediaAsset, MultimediaError]; errors are domain values, not exceptions

Testing

from lexigram import Application
from lexigram.multimedia.tts import AudioTTSModule


async def test_boot():
    async with Application.boot(modules=[AudioTTSModule.stub()]) as app:
        assert app.container is not None

Key Source Files