tts-webui-extension.openai-tts-api 1.3.2

OpenAI compatible TTS API with support for multiple TTS models

OpenAI TTS API

This extension provides an OpenAI compatible API for multiple TTS models including Kokoro TTS and RVC.

Used as extension in TTS Webui

Installation

pip install git+https://github.com/rsxdalv/extension_openai_tts_api@main

Requirements

Models require individual installation of their extensions.

# Kokoro TTS
pip install git+https://github.com/rsxdalv/extension_kokoro@main
# RVC
pip install git+https://github.com/rsxdalv/extension_rvc@main
# Chatterbox
pip install git+https://github.com/rsxdalv/extension_chatterbox@main

Usage

Authentication

You can optionally protect the API with an API key.

• Environment variable (priority): set OPENAI_API_KEY before starting the server/extension.
• UI setting: in the extension's Gradio UI, set "API Key (OpenAI-compatible)". This is saved to your config and used when the env var is not set.
• If neither is set, the API is open and does not require authentication.

Clients should send the header Authorization: Bearer <YOUR_API_KEY>.

Kokoro TTS

import asyncio

from openai import AsyncOpenAI
from openai.helpers import LocalAudioPlayer

openai = AsyncOpenAI()

async def main() -> None:
    async with openai.audio.speech.with_streaming_response.create(
        model="hexgrad/Kokoro-82M",
        voice="af_heart",
        input="Today is a wonderful day to build something people love!",
    ) as response:
        await LocalAudioPlayer().play(response)

if __name__ == "__main__":
    asyncio.run(main())

RVC

import asyncio

from openai import AsyncOpenAI
from openai.helpers import LocalAudioPlayer

openai = AsyncOpenAI()

async def main() -> None:
    async with openai.audio.speech.with_streaming_response.create(
        model="hexgrad/Kokoro-82M",
        voice="af_heart",
        input="Today is a wonderful day to build something people love!",
        extra_body={
            "params": {
                "use_gpu": True,
                "rvc_params": {
                    "pitch_up_key": "0",
                    "index_path": "CaitArcane\\added_IVF65_Flat_nprobe_1_CaitArcane_v2",
                    "pitch_collection_method": "harvest",
                    "model_path": "CaitArcane\\CaitArcane",
                    "index_rate": 0.66,
                    "filter_radius": 3,
                    "resample_sr": 0,
                    "rms_mix_rate": 1,
                    "protect": 0.33,
                },
            },
        },
    ) as response:
        await LocalAudioPlayer().play(response)

if __name__ == "__main__":
    asyncio.run(main())

Chatterbox

from openai import OpenAI

client = OpenAI(api_key="sk-1234567890", base_url="http://localhost:7778/v1")

with client.audio.speech.with_streaming_response.create(
    model="chatterbox",
    # voice="random",
    voice="voices/irohazaka.wav",
    input="Today is a wonderful day to build something people love!",
    # speed=1.0, # not implemented
    extra_body={
        "params": {
            "exaggeration": 0.5,
            "cfg_weight": 0.5,
            "temperature": 0.8,
        },
    },
) as response:
    audio = response.read()
    with open("audio.mp3", "wb") as f:
        f.write(audio)