Multi-format delivery
AI auto reframe that respects every delivery format
One video now ships as many videos: widescreen for YouTube, vertical for Shorts and TikTok, square for feeds. Cropping each by hand means re-judging every shot three times. Auto reframe tracks the subject and recomposes per format — but only earns trust when you can override its framing decisions.
From the Mellius product team · Updated September 28, 2026
Conspiracy theorists, you may hate this
AI Director
›
Just type what you need
- ›Remove all silent parts
- ›Add bold kinetic captions
- ›Only have Bill Ackman speaking
- ›Cut to highlights only
- ›Add cinematic transitions
- ›Match cuts to the beat
The short answer
AI auto reframe tracks the subject and recomposes each shot for vertical, square, or widescreen delivery without manual keyframing. Mellius keeps reframed versions as editable timeline states, so you can adjust framing per shot while preserving cuts, captions, and graphics. Always inspect faces, text, and motion at the smallest delivery size.
Subject tracking
The reframe must follow the true subject — speaker, action, or object — across cuts and movement.
Override control
Every automatic framing decision needs a fast manual correction for the shots it misreads.
Layer awareness
Captions, graphics, and lower-thirds must recompose with the picture, not strand outside the crop.
What auto reframe must get right
Tracking is the headline feature, but edge cases decide quality: two speakers in dialogue, a demonstration where hands matter more than faces, text-heavy screen recordings, fast pans that outrun the crop window. A reframe that centers faces will happily crop out the product being demonstrated. Test candidate tools on your hardest shots, not your cleanest talking head.
Motion behavior matters as much as framing. The crop window should move smoothly, ease into repositioning, and hold steady during static moments — jittery reframing reads as amateur faster than any color issue. And captions plus graphics need format-aware layouts: a lower-third designed for widescreen will collide with a vertical crop unless the system recomposes layers together.
- Test dialogue, demonstrations, screen recordings, and fast motion — not just talking heads.
- Check crop-window smoothness: no jitter, no lagging behind the subject.
- Verify captions and graphics recompose per format instead of stranding outside the crop.
Multi-format versions in Mellius
In Mellius, reframed versions stay connected to the same edit: the cuts, captions, color, sound, graphics, and masks you approved carry across aspect ratios, and framing becomes one more editable property per shot. Ask the director to recompose for vertical, adjust any shot where the crop misreads the subject, and the caption and graphic layers follow the new composition.
This beats the export-three-timelines workflow because revisions propagate. Tighten the opening or fix a caption once and every format inherits it; adjust vertical framing on one demonstration shot without touching the widescreen master. Inspect each format at its true delivery size before publishing, since readability failures hide at desktop preview scale.
Compare the workflows
| Approach | What it does well | Where it falls short |
|---|---|---|
| Single-purpose reframe tool | Fast subject tracking for one-aspect conversions | Output disconnected from captions, graphics, and revision flow |
| Mellius connected reframe | Per-format framing editable alongside the full timeline | A newer platform than the established desktop suites |
| Manual multi-timeline | Pixel-perfect composition for every format | Every revision must be repeated across all versions |
A practical sequence
Put it into practice
- 01
Lock the master
Finish cuts, captions, and pacing in widescreen before deriving formats.
- 02
Reframe and override
Auto-track each format, then correct framing on shots the tracker misreads.
- 03
Inspect at delivery size
Check faces, text, graphics, and motion on a real phone-sized preview.
Frequently asked questions
Does auto reframe work for vertical Shorts and TikTok?
Yes — vertical is the primary use case, and tracking handles centered speakers well. The failures come from demonstrations, multi-speaker scenes, and on-screen text designed for widescreen. Always override framing per shot and re-check captions and graphics inside the vertical crop.
Will auto reframe cut off faces or text?
It can, especially with multiple subjects, fast movement, or text near frame edges. Good tools let you keyframe the crop window or pin framing per shot. Inspect every scene boundary and text moment at the smallest delivery size rather than trusting the track blindly.
Should I shoot differently if I plan to reframe?
Yes. Center key action with headroom, keep text and graphics inside a centered safe area, and favor slightly wider shots than a single-format edit needs. Protection in production beats correction in post — reframe AI performs dramatically better on footage shot with multiple formats in mind.
Continue reading
Primary sources
Product capabilities change. These official references support the comparisons above.
Describe the edit. Keep the timeline.
Mellius turns conversational direction into cuts, captions, graphics, motion, color, and sound you can still inspect and change.
Start editing free