Changelog

Versions use semver (major.minor.patch). You always get the version shown in Help → About and in the Windows installer filename.

0.2.3

2026-07-24

Transcription timing: more accurate VAD and cue packing; fixes false multi-speaker lanes when one person talks
Timeline editing: edge handles use linked / snap / free zones; subtitle text stays clear in the center
Overlapping duplicate cues from segment ASR are deduped automatically

2026-07-23

Cue packing: short phrases no longer stretch across pauses; long lines split at natural breaks (and / comma)
Filters Whisper hallucination loops on silence; Delete all cues clears the auto-saved SRT; Generate refreshes that file cleanly
Timeline zoom up to 1024× for fine edits on smaller screens
Marketing site screenshots and compare / downloads visuals

2026-07-23

GPU: Generate checks CUDA runtime before transcribing; installing GPU support from Models refreshes acceleration without restart
Timing: Balanced mode uses finer alignment when the model supports it; fewer fixed ~2s cue blocks on long clips
Language: Default primary language follows your Windows display language (saved in Settings)
Overlapping speakers get distinct colors in the editor and export styling
Trial export limits, native video audio, and quieter background installs (see 0.2.0 notes where applicable)

2026-07-23

Trial: Unlimited Generate; export trial was 10 hard burn-ins (max 0:30) — see Help for current limits
Settings: View → Settings — customize keyboard shortcuts (open video, save, undo, play/pause)
Cleaner editor layout and a slimmer toolbar (project actions stay in the File menu)
Language panel: clearer input vs output; subtitle tracks can translate to your chosen output languages
Loading indicator while a video opens; stereo level meters in the player
Clearer messages when transcription fails
Check for updates uses the same version number as the app