Voice memo → text
Web: Tiny Q5_1 only. RIFF/WAV is streamed up to 1 GiB; browser-decoded compressed files up to 32 MiB, at most 5 minutes and exactly 19,200,000 bytes (19.2 MB / about 18.31 MiB) mono 16 kHz Float32 PCM. iOS: Apple-readable input up to 1 GiB and 5 minutes, protected staging in 1 MiB chunks, TXT up to 1 MiB, and a 10-minute watchdog; available Apple on-device recognition in an explicitly chosen language only, no Base/Whisper model and no network fallback. No audio upload.
ready
Processing runs on your device.
Frequently asked questions
Which recognition runs on each platform?
Tiny Q5_1 only on the web. iOS bundles no Whisper or Base model and uses only Apple on-device recognition available for an explicitly chosen language.
Which voice-memo limits apply?
Web: RIFF/WAV is streamed up to 1 GiB; browser-decodable compressed files are capped at 32 MiB, duration at 5 minutes, and decoded PCM at exactly 19,200,000 bytes (19.2 MB / about 18.31 MiB). iOS: Apple-readable input up to 1 GiB and 5 minutes, 1 MiB staging chunks, TXT up to 1 MiB, and a 10-minute watchdog.
Is audio uploaded?
No.