Local Voice Synthesis
Generate clean speech
directly in the browser
Pick a voice, paste a line, and hear the result locally.
No backend, no signup, no audio leaving the page.
Independent community port. Attribution and source links are in the footer.
25 MB
nano model
24 kHz
sample rate
8
voices
Local inference
8 voices
24 kHz output
No server needed
StyleTTS2-based
Start here
The first run may take a few seconds while the model downloads and warms up.
Browser support is split by model: Nano ONNX runs on WebGPU or WASM, while Micro and Mini ONNX stay on CPU/WASM.
1.0×
Threads
Compare cold start separately from steady-state synthesis. The throughput charts below unify Nano, Micro, and Mini in one view for the selected workload and summary statistic.
Loading benchmark data…
Loading benchmark data…