POST your text, get natural, low-latency speech back as WAV, PCM or μ-law. 15 voices, 31 languages, faster than real time — no SDK, no GPU, no fuss. Try it right here:
$ curl -X POST https://speak.mosaic.site/v1/speak -H "Authorization: Token sk_…" -d '{"text": "Hello world"}' -o hello.wav
Mosaic API keys work too — pass your mk_… key the same way.
Click any card to hear a sample. The five Pocket voices speak English; the ten Supertonic voices speak all 31 languages and understand inline expression tags like <laugh>, <sigh> and <breath>.
Drop your email, get a key instantly. No card, no verification dance. 10,000 characters per request, 60 requests a minute.
Save it now. This key is shown exactly once — we only store a hash.
Everything an agent needs — auth, endpoints, voices, limits — lives in one markdown file it can fetch and follow:
Add this to CLAUDE.md, AGENTS.md, a system prompt — anywhere your agent reads instructions: