v0.5.0 turns Mousify from a recorder into a finisher. Record, click once — subtitles, camera work, pacing and voiceover are done. All that's left is a look-through and Export..
⚡ Quick Cut: one click after recording → Full (keep everything, 2× through idle) or Reel (work moments only). Everything lands on the timeline — visible, adjustable, undoable. Your manual timeline marks always win over the algorithm.
🎙 AI Subtitles (fully on-device): local whisper transcription, sentence-level cues, voice-onset-aligned timestamps. Re-transcribe just a selected span when a few cues are off. Robust language auto-detect.
🔤 Subtitle import/export + dubbing: round-trip .srt/.vtt with frame-accurate timing; TTS-dub every cue (or re-voice just the selected ones). Local voices + BYOK cloud (DashScope / ElevenLabs) + voice cloning.
🎬 AI Narration Director: drafts a full narration from your recording (on-device signals only — frames never leave your machine), styles from tutorial to esports caster; per-segment voicing and rewrite; narration → subtitles in one click.
🖼 Canvas mode: video floats on a color/gradient/custom background with adjustable scale, corners, shadow — plus cursor-follow panning that keeps your pointer centered. Built for vertical/TikTok formats.
🔦 Spotlight: dim everything outside a circle so the viewer looks where you want — following the cursor, or pinned where you put it. Quick Cut places them for you, on the seconds where you were working in one dense spot. Every one lands on the timeline: visible, movable, deletable.
🎞 Multi-clip stitching: drop in a whole folder — mismatched resolutions, frame rates and codecs are normalized and joined in your order, then opened as one clip.
👤 AI presenter: bind an avatar to a voice track and put a face on screen, with the same shape controls as the webcam PiP.
🌐 Subtitle translation: translate existing captions into another language onto layer B — stack both, or show one.
🤖 Your AI assistant can drive the editor (MCP): Claude Code / Codex / Cursor can read your timeline, look at the composited picture and make edits — the same actions you have in the UI, each call one undo step. One-click setup in Settings.
✂ Editor upgrades: per-range + global speed with preview=export parity, event chips, AI chapters, freeze-frames, silence trimming, multi-track audio, BGM, covers, dissolve transitions, privacy scan with one-click blurring, double-click .mousify to open, a low-overhead preview copy for big footage (export always uses the original), and zero-copy full-screen capture.
🛠 Fixes: export duration now matches the timeline exactly; clean 1920×1080 output; full-frame marks correct in smart-zoom exports; honest error messages when recording over Remote Desktop; much lower memory and CPU during long previews (the playhead no longer re-renders at monitor refresh rate; scan progress and spotlight drawing are throttled), plus a memory guard that offers Save & refresh as you approach the ceiling.