Embed powerful natural language understanding into your web apps. No servers. No API keys. Zero latency.
Use our visual builder to map voice commands to actions. Define variables like {color} or {room} instantly.
We automatically generate a lightweight JSON model and TypeScript definition tailored to your intents.
Download `nanovoice.ts` and drop it into your project. One line to initialize, one event to listen.
Perfect for kiosks, subways, or anywhere with spotty internet. Your app never breaks.
No round-trip to the server. Intent recognition happens instantly on the user's device.
Switch intelligently between Online Speed (Web Speech API) and Offline Privacy (Whisper). Best of both worlds.
NanoVoice runs entirely on your hardware. We use FunctionGemma-270M for intent recognition and Whisper-Tiny for speech processing.
Utilizes WebGPU for near-zero latency on modern browsers.
Balanced for 16GB RAM systems to prevent browser overhead.
| dtype | Approx size | Browser safe |
|---|---|---|
fp32Current | ~1.1 GB | crash |
fp16 | ~550 MB | risky |
q8 | ~700 MB | maybe |
q4 | ~400 MB | safe |
* Estimates include WebGPU buffer overhead. Recommended: Chrome with 16GB+ RAM.
Touchless ordering in fast food or ticketing.
Control smart home devices locally without cloud lag.
Voice commands for spells or inventory management.
Hands-free data entry for sterile environments.
Join developers building the next generation of privacy-first, voice-native web experiences.
Get Started for Free