Auto-crop to vertical, the smart way

The best AI tool for 9:16 reframing with the subject always in frame.

LumiClip watches the source frame-by-frame and crops your landscape video to a 9:16 vertical view that follows whoever is speaking, tracks the subject on screen, and avoids cropping out lower thirds or text. No manual keyframing, no off-center heads, no losing the action halfway through a clip.

Try LumiClip Free

Score96

Score88

Score94

Centre-crop doesn't cut it

Reframing isn't a crop — it's a per-frame decision.

Most tools 'reframe' to 9:16 by cropping the centre and hoping the subject stays there. That works for a single, locked-down shot. It fails the moment a speaker walks, a guest reacts off-centre, or two hosts trade off — the heads go out of frame, the screen text gets sliced in half, and the clip looks amateur in a feed full of vertical native content.

LumiClip's reframing pipeline runs per-frame subject and active-speaker analysis. It picks the right crop for the moment — following the speaker, tracking the subject when no one's talking, splitting the frame only when voices genuinely overlap, and respecting on-screen safe zones so titles and lower thirds survive the crop.

Why creators pick LumiClip for reframing

Three things to look for in an AI 9:16 reframing tool

Not every reframer reframes the same. These three differences show up in the final clip.

Active-speaker detection

For multi-host or interview content, the vertical crop should follow whoever is actually speaking. LumiClip's active-speaker detection picks single-speaker framing when one person is talking, split-screen only on real overlap, and picture-in-picture when a guest is reacting silently — so the frame always shows the person you'd want to see.

Per-frame subject tracking

When no one is speaking — gameplay footage, a reaction shot, a product demo, a panning landscape — the reframer tracks the subject across the frame instead of locking to centre. The clip stays composed even when the action moves.

Safe-zone awareness for text and lower thirds

On-screen text, lower thirds, sponsor logos, and score overlays get detected and protected. The 9:16 crop window won't slice a chyron in half or push the host's name off the edge of a vertical clip.

How it works

From 16:9 to 9:16 in three steps

Upload once. The reframer handles the rest.

Upload your landscape source
Drop a 16:9 video — podcast, webinar, interview, gameplay capture, lecture, anything. Up to a full episode at a time. Paste a YouTube or podcast URL or upload the file directly.
AI analyses every frame
Per-frame face detection, active-speaker analysis, subject tracking, and safe-zone scanning run on the source. The reframer plans a 9:16 crop path that follows the action and protects on-screen text.
Export ready-to-publish 9:16 clips
Get vertical clips with the subject framed correctly, plus animated captions and brand templates if you want them. Native-resolution exports for TikTok, Reels, Shorts, and Stories — no manual keyframing required.

What the reframer does

The reframing pipeline, end to end

Every feature here is built to keep the subject in frame and the clip publish-ready.

Active-speaker following

Detects who is talking in any given moment and tracks the vertical crop to them. Single speaker on solo moments, split-screen only on real overlap, picture-in-picture for silent reactions.

Subject tracking when no one's talking

For gameplay, B-roll, demos, or panning shots, the reframer follows the main subject across the frame instead of locking to centre. Composition stays intact even when the action moves.

Smart safe zones for text and overlays

Lower thirds, chyrons, score graphics, and on-screen text are detected and protected. The 9:16 crop won't slice off the host's name or push a caption past the edge of the vertical frame.

Multi-platform aspect-ratio presets

Export the same source to 9:16 (TikTok, Reels, Shorts), 1:1 (Instagram feed), and 4:5 (mobile-first feed) without re-running the analysis. One pass, every format ready.

What you get

Clean reframes. Zero manual keyframing.

The practical outcomes of letting AI do the per-frame work.

Zero

Manual keyframing

No more dragging crop boxes frame-by-frame to keep the speaker in shot. The reframer does the per-frame work automatically — set the source, get the vertical.

Every

Platform covered

One reframe pass produces 9:16, 1:1, and 4:5 outputs ready for TikTok, Reels, Shorts, Instagram feed, and LinkedIn. No re-export, no re-cropping per channel.

Higher

Viewer retention

Vertical clips where the subject stays in frame keep viewers watching. Off-centre heads and chopped captions are some of the fastest scroll-past signals in short-form — the reframer makes them go away.

Feature-by-feature

More reframing capabilities worth knowing

The small details that decide whether a vertical clip looks native or amateur.

Face and head detection

Detects every face on screen and prioritises the one currently speaking. Falls back to the largest or most central face when no one is talking.

Picture-in-picture mode

For reaction moments, the listening guest appears as an inset over the speaker. Useful for interviews where the reaction adds context but the audio is one-sided.

Split-screen on real overlap

Only switches to a 50/50 split when both voices genuinely overlap. Avoids the lazy default of two heads on frame the whole clip when only one is talking.

Scene-aware cut points

Reframer respects existing cuts and scene boundaries — it won't pan a vertical crop across a hard cut, which would look jittery in feed.

Subject motion smoothing

The crop path is smoothed across frames so the vertical view glides with the subject instead of jerking from face to face.

Captions burned in

Word-by-word animated captions are rendered after the reframe, sized to fit the 9:16 frame and positioned inside the safe zone.

Brand-safe templates

Logo, handle, colour palette, and lower-third positioning are baked into the export so the vertical clip ships on-brand without a manual pass.

Batch reframing

Reframe a backlog of episodes at once. The pipeline runs in parallel on uploaded sources without you babysitting each one.

AI 9:16 reframing FAQ

A centre crop takes a fixed vertical slice out of the middle of every frame — it works only when the subject is locked in the centre for the whole clip. AI reframing analyses each frame, detects the active speaker or main subject, and plans a 9:16 crop path that follows them. When the speaker walks, switches, or a guest reacts off-centre, the vertical frame moves with them instead of cropping them out.

Yes — this is where active-speaker detection earns its keep. LumiClip identifies which host is speaking at every moment and follows them; when two voices overlap genuinely, it switches to a split frame; when one guest is reacting silently, it can use picture-in-picture. The vertical clip always shows the right person.

The reframer detects on-screen text and graphic regions and treats them as safe zones — the 9:16 crop won't slice them in half or push them past the edge of the vertical frame. Lower thirds, score overlays, and sponsor logos survive the crop.

Standard 16:9 landscape is the typical input. The reframer also handles 4:3 and ultra-wide sources — anything wider than 9:16, which it will reframe to fit. The 9:16 output is native vertical resolution suitable for TikTok, Reels, Shorts, and Stories.

Yes. A single reframing pass produces 9:16 (TikTok/Reels/Shorts), 1:1 (Instagram feed), and 4:5 (mobile-first feed) versions of the same clip — no need to re-run the analysis per channel.

Yes. LumiClip's free plan includes the same active-speaker and subject-tracking reframing pipeline as paid plans, with monthly processing minutes generous enough to reframe a full podcast episode. Paid plans raise the processing cap, output at 1080p without a watermark, and unlock team workspaces.