Here's exactly what happens between dropping in a video and downloading the results — no black box, because you shouldn't trust an editor you can't audit.
Upload a file or paste a link. Set your style before the run if you want — caption style from the visual gallery, layout, accent color, zoom, music — or let Cliptica decide. Your choices always override the AI's.
Everything below happens automatically, in sequence, on every video. Most jobs finish in minutes, not hours.
Clips arrive scored and sorted. Publish the winners as-is, open any of them in the editor to adjust, and export in 9:16 at 1080p.
| Stage | Pass | What happens |
|---|---|---|
| 01 | Transcribe | Word-level transcription with timestamps and punctuation attached to the right words — the foundation every later stage builds on. |
| 02 | Find the moments | Specialized agents — story, hook, judge — score every candidate segment for virality. Weak moments are cut from the batch, not padded into it. |
| 03 | Lock the boundaries | An eight-step pipeline snaps each clip's start and end to finished sentences within 0.10 seconds, with silence-gap fallbacks for messy audio. |
| 04 | Write the hook | Multiple headline candidates are generated, scored, and filtered through validator gates that reject filler, fragments, and AI-sounding phrasing. |
| 05 | Frame the speaker | Face detection plus a seat-lock model reframes to 9:16 and holds the subject steadily — through head turns, gestures, and cuts. |
| 06 | Caption | Your chosen style renders with active-word highlighting, two-line legibility layout, and stability logic. Emoji placed by meaning at your chosen density. |
| 07 | Add the graphics | Stat cards, charts, and callouts appear where the speech supports them, populated with the exact figures spoken — density-capped and varied. |
| 08 | Mix and grade | Music under the voice at sensible levels, sound effects placed sparingly, colors synced to your accent — then the final render. |
A clip that opens mid-word reads as amateur. Boundaries are locked to sentences, verified twice — before and after every adjustment.
Motion graphics only display figures the speaker actually said. If the transcript doesn't contain it, the graphic doesn't show it.
Captions, emoji, and graphics are placed around the subject, density-capped, and mode-aware — the person stays the point.
Your first video takes about as long as reading this page did.
Join the waitlist