Which AI voice models can your device run?
Browser speech models range from lightweight CPU downloads to large WebGPU models. This check reads the capabilities your browser makes available and shows which options are a sensible match before you download anything.
What the check looks for
Browser runtime
WebAssembly supports the smaller CPU models. WebGPU can accelerate supported models or enable heavier ones.
Memory and processing
Available memory and CPU cores help estimate whether generation should feel quick, slow, or impractical.
Local model storage
The browser needs enough storage for the selected model and any voice files you choose to keep.
Download sizes and runtime requirements are also listed on every model page so you can choose before installing.
Frequently asked questions
What is WebGPU and do I need it?
WebGPU lets a browser use a compatible graphics processor for model computation. Lightweight models can use a WebAssembly CPU path, while some larger models need WebGPU for practical generation.
Why does my Android phone fail on some models?
Mobile browsers often expose less memory, storage, and GPU functionality than desktop browsers. The checker evaluates the capabilities it can read and recommends lighter models when the larger runtimes are unavailable.
Does the check upload anything or install anything?
No. The check reads capability information already exposed by your browser. It does not download a speech model or send a hardware profile to AIVoices.
My device says 'slow' — can I still use TTS?
Yes. A slow result means generation may take longer than the finished audio. Try a smaller model such as KittenTTS or a CPU-oriented Piper voice before downloading a heavier option.