Staging · Invited testers only · Test payments
Speech, with a little more character.

Give your words
a voice.

Find a voice that feels right. Try open speech models, then bring your favorite into your product through one API.

Open Playground Build with the API
Four models. A free place to try them.
GOOD WORDS DESERVE A GOOD VOICE
In
IndexTTS-2.5English reference · Sample clip
WAV

Welcome to Timbrelay.
Give your words a voice of their own.

Hear the voice
0:04
Watermark appliedBuilt into the audio.
In the studioAuK BaseAuK FlashIndexTTS-2.5CosyVoice3
THE VOICE COLLECTION

Same words.
A different feel.

Listen, compare, and find your starting point.
These four models are connected in the development Playground.

Ab

AuK Base

The full AuK model, for everyday voice work.

24 kHz · WAVTry voice →
Af

AuK Flash

A lighter AuK option for trying out short passages.

24 kHz · WAVTry voice →
In

IndexTTS-2.5

A different character for narration and spoken text.

22.05 kHz · WAVTry voice →
Cv

CosyVoice3

A versatile choice for multilingual speech.

24 kHz · WAVTry voice →
A PLACE FOR YOUR VOICE WORK

From a quick idea
to your application.

Experiment in Playground. Keep API access organized by project. Bring the audio into whatever you are building.

Start a project

Room to try things

Change the model, voice preset, text, or pace. Your signed-in Playground is free to use.

A watermark you don’t hear

When watermarking is enabled, audio is encoded before you receive it. Each response reports its actual watermark status.

Projects with a purpose

Give each application its own API keys. Shared audio history and inviting collaborators are next.

Team features soon
GETTING STARTED

A short path to speech.

01

Find your voice

Try a model and a shared voice preset in the free Playground.

02

Make it part of your project

Create a project key and send the same request from your application.

03

Take the audio with you

Receive a complete WAV. In watermark mode, encoding finishes before delivery.

WHEN YOU ARE READY TO BUILD

A familiar request.
A voice of your choosing.

One endpoint, a project key, and a few lines of text. Switch models without rewriting your integration.

Try an API request

The program API uses your credits. Playground is free.

speech.wav
curl https://api.staging.timbrelay.com/v1/audio/speech \
  -H "Authorization: Bearer $TIMBRELAY_PROJECT_KEY" \
  -H "Idempotency-Key: first-voice-clip" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "cosyvoice3",
    "voice": "voice_preset_en_5639",
    "input": "Welcome. It is good to have you here.",
    "response_format": "wav"
  }' --output speech.wav
Local development endpoint · Complete WAV response