Home

Most editing tools cut for cleanliness: strip the filler words, tighten the gaps, done. Emotion Engine scores delivery and ranks repeated attempts using pace, pitch, and energy. Dialogue cleanup stays optional: Filler Review proposes changes, and nothing is removed until you choose it, save the edit, and re-render.

That is the whole bet: picking the best take beats picking any clean take. Every clip it hands you is already the strongest version of that moment, not just the tidiest.

On Home, drop a video (or paste a link), then hit Get clips in 1 click. The engine transcribes it, scores prosody word by word, detects repeated takes, mines the strongest moments, and renders them as horizontal clips.

Everything runs on this Mac. When it finishes, open any clip in the editor to trim, caption, and reframe it. Your footage never gets uploaded; only audio goes to the AI providers you set up with your own keys, and only during analysis.

Each project is a folder with a scored_transcript.json. The pieces that matter:

• Segments: the transcript, with word-level timings and prosody z-scores (pace, pitch, energy) for every word.
• Take groups: attempts at the same line, ranked so you can review possible retakes and restarts.
• Verbal tics: filler words and phrases flagged for optional review, never removed by default.
• Clip candidates: the moments worth posting, each with a hook score, delivery score, and topic overlap.

The app reads this file but never rewrites the engine's analysis. Your edits live in a separate user_edits.json sidecar, so you can always get back to the original cut.

Click any word to jump the player there. Struck-through words are cut. Select words to cut them, click a cut to audition or restore it. The waveform timeline, trim bar, and splice strip below all mirror the transcript, so you can work whichever way feels faster.

• Blade (B): click the timeline to create an edit point.
• Select (A): choose a physical sub-clip, press Delete to ripple-remove it, or drag a cut seam to adjust it.
• J / K / L: reverse shuttle, normal-speed play/stop, and forward shuttle. Repeated J or L presses increase speed.
• Zoom (Z) and Hand (H): get in close and pan around. Ordinary scrolling pans whenever the timeline is zoomed.
• Pen (P): add and drag volume keyframes directly over the main waveform.
• Trim bar: pull the clip's in and out points, capped at the neighboring speech.
• Volume lane: click and drag points directly, or use Add at playhead plus the Time and Level controls for precise keyboard adjustments. Playback previews the exact volume ride; exports bake it in.
• Space: play or pause from anywhere in the editor, unless you are typing in a field.
• Left / Right Arrow: move one second; hold Shift to move five seconds.
• M: mute or unmute editor monitoring. This never changes the exported audio.
• Undo / redo: ⌘Z and ⌘⇧Z. Playback updates immediately; save when the edit sounds right, then re-render only when you need a new exported MP4.

Re-render clip bakes your edit (cuts, trim, and volume) into a new MP4. It skips the AI judge, because your cut is final.

Make 9:16 vertical reframes that same cut for Reels, TikTok, and Shorts. Clips are always cut horizontal first, then reframed on demand, so the horizontal version stays untouched and you never recut for a second aspect.

In the editor's right rail, the Caption Studio previews captions on the player word by word as the clip plays. Pick a preset (Clean, Bold Pop, Minimal, Punch) or design a custom Karaoke, Pop, or Box look with the font, color, outline, position, and timing controls. Apply & burn writes a separate captioned MP4 through the deterministic engine compositor; the original stays. Keyword-specific font refinements and emoji remain separate reviewable operations.

Under Workspace › Brand Studio, set reusable caption and logo defaults. Configure project-specific Intro and Outro media under Project › Project Assets, review each real boundary in the app, then create separate branded derivatives without changing the original render.

There is no login and nothing is stored on a server. Your video stays on your machine. During analysis, audio goes to the AI providers you configured with your own keys, and nowhere else.

Analysis uses your own API keys (transcription and the clip miner). You pay those providers directly, at cost. The Total spend figure in the top bar tracks it across every project so there are no surprises.

Every project and clip has a Reveal in Finder button. Working renders stay in their project folders; Export deliverables copies MC, CH, or cleaned FV outputs into deliverables/ with client-ready names.