Holler 1.1.0

New default keys: hold Ctrl+Shift to dictate, Ctrl+Shift+L to learn a correction (no more Start menu). Spoken commands, smart formatting (25%, ₹500, emails), undo by voice ("scratch that"), safer learning, starter vocabulary packs, a redesigned Settings window with a Models page (search, download, delete, add from Hugging Face), a sleeker status pill, and reliability fixes (crash supervisor, recovery after sleep, ghost keys, verified downloads). Full list in CHANGELOG.md.

Holler 1.0.0

First release on PyPI (pip install holler). Everything from 0.3 to 0.7: background mode with tray, setup wizard and Settings, model mirrors with checksums, custom models, engine plugins, holler bench, optional Parakeet engine, Hinglish prompt support. Known issue: the status pill can stay hidden behind Claude Desktop's window on some setups.

Model mirror (Whisper small.en)

Fallback mirror of the Whisper models Holler downloads. Holler tries Hugging Face first and uses these files if that fails. Files are named <model>-<file>, e.g. small.en-model.bin. Source: Systran/faster-whisper-small.en (MIT licence).