Smart reframe to vertical

Horizontal footage cropped naively to 9:16 cuts off foreheads and slices UI text in half. Smart reframe crops with the subject in mind — faces bias upward, screen recordings fit instead of crop, products get safer framing — so one prompt turns a landscape edit into a publishable vertical.

Prompts to paste

"Reframe this to 9:16 for TikTok — keep my face in frame."

"Make it vertical, product focus — the bottle needs to stay centered."

"Vertical, but don't crop the screen recording — fit it with the UI readable."

"Reframe to portrait; the subject is on the left side of the frame."

"Turn this into a Reel: vertical, captions, tight cut."

That last one is the real workflow — reframing is one step in a delegation prompt, and the agent sequences it with everything else.

What the agent actually does

Two tools handle vertical conversion:

• Fast reframe — sets the canvas to 1080×1920 and scales every visual clip to fill (crop) or fit (letterbox) the new frame. No subject awareness.
• Subject-aware reframe — after analyzing your footage, it frames each clip using the visual analysis. Faces bias upward so eyes stay visible; screen/text clips fit instead of taking an unsafe crop; product clips get safer lower framing.

Focus modes: auto (per-clip decision, the default), face, product, screen, center, left, right. Left/right are for footage where the subject sits off-center. It reframes main-track and overlay visuals by default, or just the clips you point at.

This is heuristic framing from visual analysis, not pixel-level face tracking — the crop is a smart static choice per clip, it doesn't follow a moving subject frame-by-frame.

Fixing a crop

"The crop cut off my face around 0:30 — reframe that clip with face focus."

"Frame the demo section with screen focus instead."

Or manually: select the clip → Properties → Transform → adjust Position/Scale (Properties panel).

When to do it manually

Set the canvas in project settings, then position and scale each clip in the Properties panel. Fine for one clip; the agent is faster for a full timeline.

Limits

• Heuristic, not tracking: a subject that walks across frame mid-clip can leave the crop. Split the clip and reframe each half, or use motion tracking for follow behavior.
• Smart framing depends on visual analysis — the agent analyzes the footage first if it hasn't seen it (this counts against free-plan analysis limits, see Plans and limits).
• Reframe before captions and highlight extraction — they work best on the final canvas.
• Creator templates set the canvas and reframe as part of their flow, but captions are generated separately — a template alone doesn't transcribe anything.

See also

• Make a TikTok — full horizontal-to-vertical workflow
• Visual analysis — what the agent learns about your footage
• Creator templates — formats that include reframing