| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
| |
Pre-existing whitespace/style drift caught by `just check`. Touch
nothing semantic — pure formatter output (shfmt -i 2 -ci -s, ruff,
prettier, taplo). Excludes dot_config/clangd/config.yaml whose manual
indentation is intentionally preserved.
|
| |
|
|
|
|
|
|
|
|
| |
large-v3-turbo-q5_0 ran ~1-2x realtime on the T490's CPU, making
push-to-talk feel sluggish. The base multilingual model is ~142 MB
(vs 547 MB) and runs ~7-10x realtime, dropping perceived latency on
short utterances from a few seconds to near-instant.
Quality on short EN/PT dictation remains usable; bump WHISPER_MODEL
to small or large-v3-turbo if accuracy matters more than latency.
|
|
|
Push-to-talk dictation toggle on Super+i: parecord captures 16 kHz mono
WAV, whisper-cli transcribes (auto language), output is typed via wtype
and copied to the clipboard.
Region OCR on Super+Shift+o: slurp + grim feed tesseract (eng+por),
result lands in the clipboard with a notification preview.
Adds wtype to wayland.txt; tesseract (+eng/por data) and whisper.cpp +
the large-v3-turbo-q5_0 model package to extra.txt.
|