TTSFM: OpenAI-Compatible Text-to-Speech API Service (Project Notice)

Introdução

TTSFM was an open-source project that aimed to provide a free, OpenAI-compatible text-to-speech API service. It offered a comprehensive solution for converting text into natural-sounding speech, leveraging the openai.fm backend, which was based on OpenAI's GPT-4o mini TTS. The project included a powerful Python SDK, RESTful API endpoints, and an intuitive web playground for testing and integration.

?? NOTICE: This project is no longer functional as the openai.fm demo website has been shut down.

Key features that TTSFM offered included:

Multiple Voices: A selection of 11 OpenAI-compatible voices (alloy, ash, ballad, coral, echo, fable, nova, onyx, sage, shimmer, verse).
Flexible Audio Formats: Support for 6 audio formats (MP3, WAV, OPUS, AAC, FLAC, PCM).
Speed Control: Adjustable playback speed from 0.25x to 4.0x.
Long Text Support: Automatic text splitting and audio combining for extended content.
Real-time Streaming: WebSocket support for streaming audio generation.
Python SDK: Easy-to-use synchronous and asynchronous clients.
Web Playground: An interactive web interface for testing and experimentation.
Docker Ready: Pre-built Docker images for instant deployment.
OpenAI Compatible: Designed as a drop-in replacement for OpenAI's TTS API.

Instalação

Although the project is no longer functional, for historical and educational purposes, here were the installation methods:

Python package

pip install ttsfm        # core client
pip install ttsfm[web]   # core client + web/server dependencies

Docker image

TTSFM offered two Docker image variants:

Full variant (recommended)

docker run -p 8000:8000 dbcccc/ttsfm:latest

This variant included ffmpeg for advanced features like all 6 audio formats, speed adjustment, and format conversion.

Slim variant - ~100MB

docker run -p 8000:8000 dbcccc/ttsfm:slim

This minimal image provided basic TTS functionality with MP3 and WAV formats only, without speed adjustment or advanced conversion.

The container exposed the web playground at http://localhost:8000 and an OpenAI-compatible endpoint at /v1/audio/speech.

Exemplos

Here are examples of how TTSFM could be used:

Python client

from ttsfm import TTSClient, AudioFormat, Voice

client = TTSClient()

# Basic usage
response = client.generate_speech(
    text="Hello from TTSFM!",
    voice=Voice.ALLOY,
    response_format=AudioFormat.MP3,
)
response.save_to_file("hello")  # -> hello.mp3

# With speed adjustment (requires ffmpeg)
response = client.generate_speech(
    text="This will be faster!",
    voice=Voice.NOVA,
    response_format=AudioFormat.MP3,
    speed=1.5,  # 1.5x speed (0.25 - 4.0)
)
response.save_to_file("fast")  # -> fast.mp3

CLI

ttsfm "Hello, world" --voice nova --format mp3 --output hello.mp3

REST API (OpenAI-compatible)

# Basic request
curl -X POST http://localhost:8000/v1/audio/speech \
  -H "Content-Type: application/json" \
  -d '{
    "model": "tts-1",
    "input": "Hello world!",
    "voice": "alloy",
    "response_format": "mp3"
  }' --output speech.mp3

# With speed adjustment (requires full image)
curl -X POST http://localhost:8000/v1/audio/speech \
  -H "Content-Type: application/json" \
  -d '{
    "model": "tts-1",
    "input": "Hello world!",
    "voice": "alloy",
    "response_format": "mp3",
    "speed": 1.5
  }' --output speech_fast.mp3

Porquê usar (Historical Context)

TTSFM was a significant project for several reasons, even though it is no longer active. It demonstrated how an OpenAI-compatible text-to-speech service could be self-hosted and offered for free, providing an alternative for developers. Its comprehensive feature set, including multiple voices, audio formats, speed control, and long text support, made it a versatile tool for various applications. The project's Python SDK and Docker readiness also highlighted its ease of integration and deployment.

?? Disclaimer: This project was intended for educational and research purposes only. It was a reverse-engineered implementation of the openai.fm service and was not recommended for commercial use or in production environments. Users were responsible for ensuring compliance with applicable laws and terms of service.