 ciregenzandClaude Fable 5
|
3c24c10cbd
|
[eric] voice: prefetch the dictation model at boot so the first press never waits (ENG-380)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01En8dRGsJPLrJCQBEkTH4Mp
|
2026-08-20 22:22:48 -07:00 |
|
|
ciregenz
|
824bc10566
|
[eric] voice: boot-warm whisper only on installs that have actually dictated; a session open still warms it mid-speech
|
2026-08-16 20:07:37 -07:00 |
|
|
ciregenz
|
ef594f737d
|
[eric] voice: survey-grade accuracy layer, Chromium AGC off, suppress-nst+no-context+flash-attn+temp ladder, window energy gate, sub-second zero-pad, stutter and glossary-echo filters
|
2026-08-06 00:22:32 -07:00 |
|
|
ciregenz
|
5130309a72
|
[eric] voice: text-landed chime cut, haptics toggle, full-clip decode wins at stop, beam search 5; dock tile says Browsers
|
2026-08-06 00:02:52 -07:00 |
|
|
ciregenz
|
110d5a9b3b
|
[eric] voice+canvas: personal dictionary rides every decode, local dictation history with copy-back, double-click empty canvas fits everything
|
2026-08-05 22:43:15 -07:00 |
|
|
ciregenz
|
ba8a9c4038
|
[eric] voice: scratch-that discards the take, multilingual Small with auto language detect joins the catalog
|
2026-08-05 21:19:00 -07:00 |
|
|
ciregenz
|
f75ae3fbf7
|
[eric] voice: small.en-q5_1 is the accuracy-first default, bundled base serves while it downloads in background
|
2026-08-05 20:17:27 -07:00 |
|
|
ciregenz
|
d9813de3cd
|
[eric] voice: whisper's non-speech captions no longer inject as dictation (ENG-167)
|
2026-08-05 18:07:23 -07:00 |
|
|
ciregenz
|
2e0572ca72
|
[eric] voice: whisper boots from a private empty cwd and reaps mid-boot children, the temp-dir scan wedge killed dictation
|
2026-08-04 16:43:21 -07:00 |
|
|
ciregenz
|
26e2a8e6d7
|
[eric] voice: streaming dictation engine, phrases commit once and preview re-decodes only the open phrase
|
2026-08-04 00:10:53 -07:00 |
|
|
ciregenz
|
2572f9ff45
|
[eric] voice: label the model catalog with the latencies actually measured, not guesses
|
2026-07-31 00:10:03 -07:00 |
|
|
ciregenz
|
bce7e1f072
|
[eric] voice: Settings model picker with verified downloads; report fast whisper exits
|
2026-07-31 00:03:27 -07:00 |
|
|
ciregenz
|
664608b4a8
|
[eric] voice: warm whisper at boot, unload when idle, and stop guessing the server port
|
2026-07-30 21:49:16 -07:00 |
|
|
ciregenz
|
e321a35894
|
[eric] voice: drop whisper --convert (demanded ffmpeg on PATH at boot; Finder apps have none = dictation dead in prod), writable cwd, drain stderr, 60s cold-load budget
|
2026-07-29 16:51:30 -07:00 |
|
|
ciregenz
|
17364d5439
|
[eric] voice: fix stuck model-downloading state + harden download + visible status overlay
|
2026-07-21 23:59:47 -07:00 |
|
|
ciregenz
|
237e8f90b2
|
[eric] voice: wire both mics (help pill + spawn composer) to one shared recorder + first-run model download
|
2026-07-21 23:35:20 -07:00 |
|
|
ciregenz
|
7baba619cf
|
[eric] voice: local whisper.cpp dictation engine (warm whisper-server + transcribe/inject/hotkey IPC)
|
2026-07-21 23:06:58 -07:00 |
|