CrisperWhisper 2.0 Turbo

High-accuracy verbatim & intended speech recognition with word-level timestamps. The model captures every filler, stutter, and false start (verbatim mode) or produces a clean, readable transcript (intended mode).

Model Card | GitHub