Voice Models
VocalLab offers four speech models. They use the same voices — what changes is expressiveness, language coverage, speed, and cost. You pick the model in the generator (Voice settings → Model), and it applies to that generation.
The four models
| Model | Best for | Languages | Steerable | Point cost |
|---|---|---|---|---|
| VocalLab Studio (default) | Expressive, directed delivery; less-common languages | 200+ | ✅ Yes | 1× |
| VocalLab Studio Flash | High volume and fast turnaround, same language reach | 200+ | ❌ No | ¾ |
| VocalLab Pro | The highest-fidelity read in the core languages | 15 | ❌ No | 1× |
| VocalLab Lite | Fast drafts, bulk and long-form on a budget | 15 | ❌ No | ½ |
- VocalLab Studio — our newest model and the default on every plan, Free included. It's steerable: it follows plain-English performance instructions like
[say quietly in a low tone]and non-verbal tags like[laugh](see Voice Steering). It also covers 200+ languages and locales, so it's the model to use for anything outside the core 15. - VocalLab Studio Flash — the same generation of model as Studio, tuned for speed: about 5× faster to start speaking, at ¾ of the points. It shares Studio's 200+ languages and works with instant voice cloning, and it still reads non-verbal tags like
[laugh]. What it does not do is follow steering instructions — for a directed performance, use Studio. - VocalLab Pro — tuned for the most natural, highest-fidelity read across the 15 core languages. It doesn't follow steering instructions — those are read literally, so use Studio when you want to direct the performance.
- VocalLab Lite — a lighter, faster model that costs half the points. Great for first drafts, high-volume jobs, and long-form audio where budget matters.
Language coverage
All four models share the same 15 core languages at native-speaker quality:
Arabic, Chinese, Dutch, English, French, German, Hebrew, Hindi, Italian, Japanese, Korean, Polish, Portuguese, Russian, Spanish.
- VocalLab Pro and VocalLab Lite support these 15 core languages.
- VocalLab Studio and VocalLab Studio Flash support the 15 core languages plus 90+ additional experimental languages and 200+ regional locales — so it's the model to reach for anything outside the core 15. Experimental languages are a research preview: quality is good but can be less consistent than the core 15.
Voice cloning covers the same set. Pick the language that matches your audio sample — every language can be selected by name — or choose Auto-detect from sample and the model works it out from your recording.
Which should I use?
- Want emotion, emphasis, or a specific delivery? → VocalLab Studio.
- Generating in a language not in the core 15? → VocalLab Studio (only it covers 200+).
- Generating a lot in a non-core language, or want it back fast? → VocalLab Studio Flash.
- Want the cleanest, most natural default read? → VocalLab Pro.
- Generating a lot, or drafting, and want to save points? → VocalLab Lite.
Availability & cost
- All four models are available on every plan, Free included. VocalLab Studio is selected by default; switch in Voice settings → Model whenever you like.
- VocalLab Studio Flash bills at ¾ and VocalLab Lite at half the points of Studio and Pro. Cost is always based on text length — see Credits & Minutes.
In the API & AI tools
The API and the MCP server can use all four models (API and MCP access are included on the Pro and Max plans). Pass a model on POST /api/v1/tts — one of v-studio, v-flash, v-pro, or v-lite — or call GET /api/v1/models to list them. When you omit model, the API uses v-pro, so existing integrations keep the voice they were built on; pass v-studio to use the app's default. See the API Reference and MCP Server.