NewGPT-6.2 Astra, GPT Image 2.6, Kling V4 and Kimi K3.1 Pro are now available.

Tchavi
ModelsStudioPricingDocsBlog
Tchavi
  • Why Tchavi
  • Models
  • Pricing

Tchavi is an OpenAI-compatible AI API: 100+ models, from text to video, topped up with mobile money in FCFA or your local currency, no bank card needed.

  • support@tchavi.com
  • Support

Popular models

  • GPT-6 Sol API
  • Claude Sonnet 5.5 API
  • Claude Opus 5.5 API
  • Gemini 4 Argon API
  • DeepSeek V4.1 Flash API
  • Nano Banana 2.1 (images)
  • GPT Image 2 (images)
  • Wan 3.0 (video)
  • All 100+ models

Developers

  • Quickstart
  • API Reference
  • Claude Code
  • Cursor
  • SDKs

Resources

  • Documentation
  • Blog
  • Image models
  • Video models
  • Music models
  • Audio models
  • Mobile money payments
  • Pay for AI APIs with M-Pesa
  • Claude Code with mobile money

Legal

  • Privacy
  • Terms of Service

APIs by creator

  • OpenAI API
  • Gemini API
  • Claude API
  • Qwen API
  • DeepSeek API
  • Seedance API
  • MiniMax API
  • Mistral API

Mobile money payments · Available in 12 countries: Benin, Cameroon, Congo, Côte d'Ivoire, DR Congo, Gabon, Kenya, Rwanda, Senegal, Sierra Leone, Uganda, and Zambia. Operators: Moov Money, MTN MoMo, Airtel Money, Orange Money, M-Pesa, Free Money, and Zamtel Kwacha.

© 2026 Tchavi. All rights reserved.

We use cookies to enhance your experience, provide personalised content, and analyse our traffic; see our Cookie Policy.

Back to models
G

Lyria 3.5

google-deepmind/lyria-3.5

A whole song from a text prompt, 58 F a track

Lyria 3.5, from Google DeepMind, writes a complete song from a text prompt, up to about three minutes long. The vocals follow the language of your prompt. Google's own example is in French, and it publishes no list of supported languages. There is no separate lyrics or length field. You paste your own lyrics into the prompt, mark sections with tags such as [Verse] and [Chorus] or with timestamps, and set the length in words, so the result is not exact. It does not edit or extend an existing song. Output is MP3 or WAV, stereo. Google says every file carries an inaudible SynthID watermark that marks it as AI-generated.

Modalities
Rate limit
30 RPM

Last updated Nov 19, 2026

What this model does well

  • 💿

    58 F a song

    One flat price per song, up to about three minutes

  • ✍️

    Your lyrics, tagged

    Paste your own lyrics into the prompt and mark sections with [Verse], [Chorus] or timestamps

  • 🗣️

    Sung in your language

    The vocals follow the language of your prompt; Google's own example is in French

  • 🎧

    MP3 or WAV, watermarked

    MP3 or WAV in stereo, with an inaudible SynthID watermark that marks the file as AI-generated, per Google

Example prompts

  • [Verse] The sun rises over Grand-Popo, the pirogues come home one by one. [Chorus] We move on, we move on, together. Soft afrobeats, highlife guitar, a warm voice singing in French.

  • Festive coupé-décalé for a drinks advert in Cotonou: brass, percussion and a tagline sung in French.

  • West African gospel, four-part choir, organ and drums, hopeful lyrics in French, the feel of a Sunday morning in Ouidah.

  • Slow highlife for the opening of a podcast about the fishermen of Grand-Popo: clean guitar, soft brass, a voice singing in French.

Good to know

  • You cannot rely on the length you ask for. We asked for exactly 20 seconds and got 62.589 seconds, 3.1 times too long. The model kept our line, then invented a second line and a four-line chorus we never wrote. There is no duration field: length is steered only by words in the prompt, and in this test those words did not hold. This is one run, so it shows what can happen, not how often. If you need a jingle of a precise length, test before you depend on it.
  • You pay per song, not per second, so a short idea costs the same as a long one. Prompt carefully.
  • It does not edit or extend an existing song. Each generation is a new song.
  • Nobody can guarantee a song is free of resemblance to existing music. Check before you publish commercially.
Video and music take time to make. We keep the text you send, so the request can be followed. Your reference files and the result are deleted after 24 hours, so download yours before then.

Technical specifications

Provider
Google DeepMind
API identifier
lyria-3.5
Category
Audio
Rate limit
30 RPM
Billing unit
track

Ideal for

  • A sung theme for a video or a podcast
  • A jingle idea for a shop or a radio spot
  • A song built around lyrics you wrote yourself
  • Sketching a song before booking a studio

Ready to try it?

Test this model in the playground, no setup needed.

Frequently asked questions

Can I pay with mobile money, without a bank card?

Yes. Top up your Tchavi account with Orange Money, MTN MoMo, M-Pesa or another mobile money operator. No bank card is needed. Credits land as soon as the payment is confirmed and are deducted as you use them.

Is the API compatible with OpenAI's?

Yes. Keep your current OpenAI SDK, just change the base URL to Tchavi's (shown in the API tab) and use your Tchavi key. The rest of your code stays the same.

API

import Tchavi from '@tchavi/sdk';

const client = new Tchavi({ apiKey: 'API_KEY' });

const job = await client.music.generations.create({
  "model": "lyria-3.5",
  "prompt": "Afrobeats, warm bass, talking drum, 100 BPM, a chorus that stays with you"
});

console.log('Job id:', job.id);
console.log('Credits used:', job.tchavi.credits_used);

// Poll for completion
let result = await client.jobs.retrieve(job.id);
while (result.status === 'pending' || result.status === 'processing') {
  await new Promise((r) => setTimeout(r, 2000));
  result = await client.jobs.retrieve(job.id);
}
console.log('Track URL:', result.output?.audio_url);

API parameters

Every parameter this model accepts on the API. Check the SDK for full typings.

modelRequired
string

TTS model (e.g. `lyria-3.5`).

inputRequired
string

The text to synthesize.

voiceRequired
string

Voice timbre: `alloy`, `ash`, `ballad`, `cedar`, `coral`, `echo`, `fable`, `marin`, `nova`, `onyx`, `sage`, or `shimmer`.

response_format
"mp3" | "opus" | "aac" | "flac" | "wav" | "pcm"
Default: "mp3"

Audio container format.

Pricing

per track
175.596 credits

How billing works

  • Billed per generated track, whatever its length.
  • A typical generation costs about —, or 175.596 credits.
Per day
Per week
Per month
A 10 000 F CFA top-up gives you about ~170 tracks. Pay with Orange Money, MTN MoMo or M-Pesa, no Visa card needed.

Ready to use Lyria 3.5?

Create your accountAPI Documentation

Bonus credits on your first top-up · No card to sign up