hjLabs AI Playground
Starting the on-device runtime…
Moonshine Tiny is a 27M-parameter ONNX model that transcribes speech from an audio file or a live microphone into text with timestamps. On this page it runs entirely inside your browser: the 55 MB download comes straight from the Hugging Face repository “onnx-community/moonshine-tiny-ONNX” into your browser cache, and every inference after that executes on your own GPU through WebGPU. At this size the download is over before you have finished reading this paragraph, even on a phone tethered to a hotspot, and it costs almost nothing to keep cached alongside a dozen other models. It is an English-language model; expect quality to fall off sharply on other languages.
Built for short clips: it scales compute with audio length instead of padding to 30 s, so quick utterances finish far sooner than Whisper. In practice people reach for a model like this for transcribing interviews, meetings, voice notes and video soundtracks without uploading the audio anywhere.
Available precisions: fp32 (full precision) and q8 (8-bit). fp32 comes first because at this size full precision is both the fastest WebGPU path and the best quality; the quantized variant exists for the WebAssembly fallback and for tight cache budgets. Hugging Face records about 10k downloads a month. It is a specialist choice rather than a default, which is exactly why it is worth trying when the popular model gets your particular input wrong.
Because there is no inference server behind this page, there is nothing to rate-limit and nothing to bill: the only cost of running Moonshine Tiny here is the one-off 55 MB download, and once that is cached the page keeps working with the network switched off.
moonshine tinymoonshine tiny onnxmoonshine tiny browserrun moonshine tiny locallyonnx-community/moonshine-tiny-onnxspeech to text modelmoonshine tiny webgpumoonshine tiny download sizefree speech to text modeltransformers.js model
About 55 MB for the fp32 build, including the tokenizer and config files the runtime also has to fetch — this is the whole payload, not just the weights file. It is downloaded once and then served from the browser Cache Storage API.
Yes, through the WebAssembly fallback, which is slower but works anywhere. WebGPU (Chrome or Edge 113+, Safari 26+) is what makes it feel interactive. On WASM, prefer the smaller quantized build.
No. onnx-community/moonshine-tiny-ONNX is a public repository, so your browser fetches the ONNX files directly over HTTPS. There is no account, no key and no server in between.
No. The model weights are downloaded from Hugging Face to your browser once, and every inference after that runs on your own GPU through WebGPU. Your text, images and audio are never sent anywhere — hjLabs.in has no inference server and no way to see your input.