How long does it take?

There are two passes. Face analysis samples frames at your chosen interval, roughly tens of seconds for a one-minute clip at 0.25 s. Export draws and encodes every frame, about 1–3 seconds per second of video at 720p. Trim long videos to the part you need first.

What if there are no faces, or they aren’t detected?

Stretches without a detected face either hold the last position or drift back to centre, as you choose. For landscapes, products or screen recordings, use Fixed centre or Manual keyframes only and drag the crop box at the moments that matter; the camera moves smoothly between keyframes.

The camera is shaky or too slow. What should I change?

If it feels shaky, raise Smoothness and the dead zone so small movements are ignored. If it lags or people leave the frame, lower Smoothness, raise the maximum pan speed or add look-ahead so the camera starts moving earlier. If shot changes produce slow pans, enable cutting on shot changes.

What does the blurred background mode do?

Instead of cropping the sides away, it places all or part of the original frame in the middle and fills the top and bottom with a blurred, enlarged copy of the same video, so nothing important is lost. The foreground width slider sets how much of the frame is kept: 0% is a full crop, 100% the whole original frame.

🤳

🤳Auto Reframe Video - Landscape to Vertical, Face Tracking

Turning a landscape interview, vlog, stream or lecture into a vertical video for Reels, Shorts or TikTok is tedious when people keep walking out of frame. This tool analyses the video at a set interval with the YuNet face detector, then plans a virtual camera: it frames the largest face, keeps two people together when they fit, and sticks with the current subject instead of jumping when someone briefly looks smaller or is blocked. Detections pass through a 3-point median de-spike and an α-β (simplified Kalman) filter, then a dead zone, a critically damped spring and a speed limit turn them into steady, overshoot-free pans that never leave the frame. On shot changes it can cut straight to the new position instead of panning, and when faces are lost it can hold or drift back to centre. Add manual keyframes on the timeline or drag the crop box in the source view to lock the framing for a while, or switch to fixed centre or keyframes only. Output 9:16, 1:1 or 4:5 with optional zoom, cropped to fill or as the full frame over a blurred background. The preview shows the crop box and the framing path; export draws every frame, encodes H.264 in chunks and keeps the original audio. Everything runs in your browser; your video is never uploaded.

Free · No sign-up · Runs in your browserLast updated: 2026-10-26

Loading tool...

How to use Auto Reframe Video

1Choose a landscape video and set the target ratio (9:16, 1:1 or 4:5) and fill mode
2Click Analyse faces, then fine-tune smoothness and dead zone or add manual keyframes
3Check the preview, choose resolution and frame rate, and create the vertical video

FAQ

How long does it take?: There are two passes. Face analysis samples frames at your chosen interval, roughly tens of seconds for a one-minute clip at 0.25 s. Export draws and encodes every frame, about 1–3 seconds per second of video at 720p. Trim long videos to the part you need first.
What if there are no faces, or they aren’t detected?: Stretches without a detected face either hold the last position or drift back to centre, as you choose. For landscapes, products or screen recordings, use Fixed centre or Manual keyframes only and drag the crop box at the moments that matter; the camera moves smoothly between keyframes.
The camera is shaky or too slow. What should I change?: If it feels shaky, raise Smoothness and the dead zone so small movements are ignored. If it lags or people leave the frame, lower Smoothness, raise the maximum pan speed or add look-ahead so the camera starts moving earlier. If shot changes produce slow pans, enable cutting on shot changes.
What does the blurred background mode do?: Instead of cropping the sides away, it places all or part of the original frame in the middle and fills the top and bottom with a blurred, enlarged copy of the same video, so nothing important is lost. The foreground width slider sets how much of the frame is kept: 0% is a full crop, 100% the whole original frame.