Wave
kokoro voice studio
Engine Kokoro v1
Quality Q8
Runtime —
Match checking…
| Length | CPU | GPU |
|---|---|---|
| ~300 chars | ~15s | ~3s |
| ~800 chars | ~45s | ~10s |
| ~2000 chars | ~3 min | ~30s |
127.0.0.1:7860.
*emphasis*Word renders 22% slower[voice:af_heart]Switch voice mid-text[happy] [sad] [whisper]Per-chunk emotion<break time="500ms"/>SSML silence<emphasis>word</emphasis>SSML stress<prosody rate="slow">SSML rateDrop a 5–15 second voice clip and we'll fingerprint it against every Kokoro voice on-device, then propose a 2-voice blend that gets you closest. No upload, no account.
For now, use the Match panel inside Studio — same engine, just nested.
A searchable grid of every voice you've saved. Filter by gender, tag by use-case, and rate on 8 acoustic axes — the library learns what you mean by "warm" or "soft".
Your saved voices live in the Voice Source dropdowns today — the library will surface them with previews and ratings.