whistle
Speech-to-text with Cactus Whistle: a 17 MB model running as WebAssembly in a worker. The audio never leaves this tab, and after the first load it works offline too.
status: loading model (17 MB)…
who said what, offline
keywords (one per line, biases the decoder)
00:00.0
Space starts and stops. live transcribes while you speak, clip after stop. [ speakers ] labels who talks once a recording or file is done; click a name to rename it.
or drop audio anywhere
Dropped and opened files are decoded right here in the browser, nothing gets uploaded. Transcripts stay in this browser until you clear them.