dattm24/edge-tts-fastapi

By dattm24

•Updated 20 days ago

Edge tts running on fastapi with swagger

Image

Machine learning & AI

418

Overview Tags

dattm24/edge-tts-fastapi repository overview

⁠edge-tts-fastapi

A free, OpenAI-compatible text-to-speech API backed by Microsoft Edge Neural TTS. Drop-in replacement for applications using OpenAI's TTS endpoint.

Website & Documentation⁠

⁠Features

OpenAI Compatible API — Works with any client that supports OpenAI's /v1/audio/speech endpoint
Free Azure Neural TTS — Powered by Microsoft Edge's edge-tts library, no API keys or credits needed
11 Vocal Presets — alloy, ash, ballad, coral, echo, fable, nova, onyx, sage, shimmer, verse
6 Audio Formats — mp3, opus, aac, flac, wav, pcm
SSE Streaming — Real-time audio streaming via Server-Sent Events
Speed Control — Adjust playback speed from 0.5x to 2.0x
Expanded API Support — ElevenLabs and Azure Cognitive Services compatible endpoints
Text Preprocessing — Automatic markdown, emoji, and HTML cleanup for natural-sounding output
Voice Listing — Query available voices by language/locale

⁠Quick Start

⁠Docker

docker run -d \
  -p 5050:5050 \
  -e DEFAULT_VOICE=en-US-AriaNeural \
  -e DEFAULT_RESPONSE_FORMAT=mp3 \
  -e DEFAULT_SPEED=1.0 \
  --name edge-tts-fastapi \
  dattm24/edge-tts-fastapi

⁠Install FFmpeg (optional, for non-mp3 formats)

docker run -d \
  -p 5050:5050 \
  -e INSTALL_FFMPEG_ARG=true \
  --name edge-tts-fastapi \
  dattm24/edge-tts-fastapi

⁠Docker Compose

services:
  tts:
    image: dattm24/edge-tts-fastapi
    container_name: edge-tts-fastapi
    restart: unless-stopped
    ports:
      - "5050:5050"
    environment:
      DEFAULT_VOICE: en-US-AriaNeural
      DEFAULT_RESPONSE_FORMAT: mp3
      DEFAULT_SPEED: 1.0

⁠API Reference

⁠Generate Speech

POST /v1/audio/speech

Generate audio from text. Accepts JSON body.

Request Body:

{
  "model": "tts-1",
  "input": "Hello, world!",
  "voice": "nova",
  "response_format": "mp3",
  "speed": 1.0
}

Parameter	Type	Default	Description
`model`	string	`tts-1`	TTS model name
`input`	string	(required)	Text to synthesize
`voice`	string	`en-US-AvaNeural`	Voice ID or Azure Neural Voice name
`response_format`	string	`mp3`	Output audio format
`speed`	number	`1.0`	Playback speed (0.5 – 2.0)

Available Voices (Preset IDs):

Preset ID	Azure Neural Voice	Language
`alloy`	en-US-JennyNeural	English (US, Female)
`ash`	en-US-AndrewNeural	English (US, Male)
`ballad`	en-GB-ThomasNeural	English (GB, Male)
`coral`	en-AU-NatashaNeural	English (AU, Female)
`echo`	en-US-GuyNeural	English (US, Male)
`fable`	en-GB-SoniaNeural	English (GB, Female)
`nova`	en-US-AriaNeural	English (US, Female)
`onyx`	en-US-EricNeural	English (US, Male)
`sage`	en-US-JennyNeural	English (US, Female)
`shimmer`	en-US-EmmaNeural	English (US, Female)
`verse`	en-US-BrianNeural	English (US, Male)

Or use any Azure Neural Voice name directly (e.g. zh-CN-XiaoxiaoNeural).

Response: Binary audio file with Content-Type matching the requested format.

cURL Example:

curl -X POST http://localhost:5050/v1/audio/speech \
  -H "Content-Type: application/json" \
  -d '{"input": "Hello, world!", "voice": "nova", "response_format": "mp3"}' \
  --output speech.mp3

OpenAI Python Client Example:

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:5050/v1",
    api_key="fake-key"
)

response = client.audio.speech.create(
    model="tts-1",
    voice="nova",
    input="Hello, world!",
    response_format="mp3",
    speed=1.0
)

response.stream_to_file("output.mp3")

⁠SSE Streaming

Add "stream_format": "sse" to request body for real-time streaming:

curl -X POST http://localhost:5050/v1/audio/speech \
  -H "Content-Type: application/json" \
  -d '{"input": "Hello, world!", "voice": "nova", "stream_format": "sse"}'

Response format (SSE events):

{"type": "speech.audio.delta", "audio": "base64EncodedChunk"}
{"type": "speech.audio.delta", "audio": "base64EncodedChunk"}
{"type": "speech.audio.done", "usage": {"input_tokens": 3}}

⁠List Available Models

GET /v1/models

Response:

{
  "models": [
    {"id": "tts-1", "name": "Text-to-speech v1"},
    {"id": "tts-1-hd", "name": "Text-to-speech v1 HD"},
    {"id": "gpt-4o-mini-tts", "name": "GPT-4o mini TTS"}
  ]
}

⁠List Voices

GET /v1/voices
GET /v1/voices/all

Returns available voices filtered by optional ?language=en-US query parameter.

⁠Expanded API Endpoints

Compatible with third-party client integrations. Requires EXPAND_API=True.

Endpoint	Method	Description
`/elevenlabs/v1/text-to-speech/{voice_id}`	POST	ElevenLabs-style TTS (expects `text` field)
`/azure/cognitiveservices/v1`	POST	Azure SSML-style TTS (expects SSML body)
`/v1/audio/speech`, `/audio/speech`	POST	OpenAI-compatible speech generation

⁠Environment Variables

Variable	Type	Default	Description
`PORT`	int	`5050`	Server port
`HOST`	string	`0.0.0.0`	Bind address
`DEFAULT_VOICE`	string	`en-US-AvaNeural`	Default voice
`DEFAULT_RESPONSE_FORMAT`	string	`mp3`	Output audio format
`DEFAULT_SPEED`	float	`1.0`	Playback speed
`DEFAULT_LANGUAGE`	string	`en-US`	Default language for voice listing
`REQUIRE_API_KEY`	bool	`False`	Require `X-API-Key` header
`REMOVE_FILTER`	bool	`False`	Skip markdown/emoji text preprocessing
`EXPAND_API`	bool	`True`	Enable ElevenLabs & Azure endpoints
`DETAILED_ERROR_LOGGING`	bool	`True`	Verbose error output in logs

⁠License

[Add your license here]

Built with edge-tts⁠, FastAPI⁠, and ❤️.

Website & Documentation⁠

Tag summary

Recent tags

Content type

Image

Digest

sha256:f95b68165…

Size

78.9 MB

Last updated

20 days ago

docker pull dattm24/edge-tts-fastapi