Version 1.3.24

[maboa]

• Whisper Small and Large v3 Turbo are hidden from the local model menu: transcribing with Small froze a Mac outright, most likely the GPU memory growth tracked in Whisper WebGPU VRAM usage grows with media length — 30-min file exhausts 12 GB on small #462. A saved choice of either falls back to Base.
• When a local transcription finishes in a background tab, the transcript appears straight away; the count up to 100% could hold it back for over a minute (Local transcription progress moves in ~13% steps: report it within each window #676 follow-up).

Refs #462