Auto Reframe Video: AI Horizontal to Vertical That Follows Faces
AIAuto reframe video from landscape to vertical without losing the speaker. A face detector follows the main person through the clip and a 9:16, 4:5 or 1:1 crop glides with them like a calm camera operator, instead of a fixed centre crop that cuts people off. It runs in your browser: no upload, no watermark and no editor to learn.
- Runs on your device
- On-device AI
- WebCodecs · GPU
- Works offline
Updated
Drop your video here
How to auto reframe a video to vertical
Step 1: Drop a landscape video on the box above: an interview, a podcast, a talk or a vlog.
Step 2: Choose the Target shape (9:16, 4:5 or 1:1) and keep Follow on Faces.
Step 3: Set Camera smoothness; the default 70 % moves gently. Then press Reframe.
Step 4: Preview the vertical video and download it.
Good to know
Settings explained
- Camera smoothness
- Higher feels like a calm camera operator.
- Quality
- Only used when a stream has to be re-encoded. Compatible streams are copied with no quality loss.
- Faster: Smaller file, quickest export
- Balanced: Good quality at a sensible size
- Best quality: Largest file, slowest export
Formats
How it runs
- Vision model
- A compact computer-vision model runs locally in your browser. Nothing is sent to an AI service.
- Hardware video engine
- Video is decoded and encoded by your device’s media hardware through WebCodecs — usually several times faster than real time.
- GPU canvas
- Pixels are drawn and composited with the GPU-accelerated canvas.
| Use | Shape | Pixels |
|---|---|---|
| TikTok | 9:16 | 1080 × 1920 |
| Instagram Reels | 9:16 | 1080 × 1920 |
| YouTube Shorts | 9:16 | 1080 × 1920 |
| Instagram / Facebook Story | 9:16 | 1080 × 1920 |
| Instagram feed (portrait) | 4:5 | 1080 × 1350 |
| Instagram feed (square) | 1:1 | 1080 × 1080 |
| YouTube (1080p) | 16:9 | 1920 × 1080 |
| X / Twitter | 16:9 | 1280 × 720 |
Checked 11 October 2026. Platforms change their limits — check the official help pages if a number looks different.
What is auto reframe?
Auto reframe turns a wide video into a tall or square one by moving the crop to keep the subject in view, the feature Premiere Pro and CapCut call by the same name. Here it follows faces: the largest face is the subject, and when another face becomes clearly bigger, for example the guest leaning in, the crop glides over to them.
| Source video | 9:16 | 4:5 | 1:1 |
|---|---|---|---|
| 1280×720 (720p) | 406×720 | 576×720 | 720×720 |
| 1920×1080 (1080p) | 608×1080 | 864×1080 | 1080×1080 |
| 3840×2160 (4K) | 1216×2160 | 1728×2160 | 2160×2160 |
How the face tracking crop moves
The video is scanned first, on frames scaled to at most 640 pixels wide: 5 times a second on Fast, 10 on Balanced and 15 on Best (Advanced settings). Faces are tracked between scans, and the camera path is then smoothed forwards and backwards, so the crop does not lag behind the speaker or jitter when they nod.
Camera smoothness sets how calm that path is: higher values add a larger dead zone, where small movements are ignored, and a lower top speed. If no face is found at all, the video gets a centre crop, the same as choosing Follow: Centre.
What if there are two people in the shot?
The crop follows one person at a time: the largest face, usually the one closest to the camera. When the other person’s face becomes clearly larger, the crop pans across to them. For a two-person podcast filmed wide, where both stay the same size, the crop stays with one speaker; use 4:5 or 1:1, which keep more of the frame, or the blurred-background layout of the horizontal to vertical video converter to show both.
Auto reframe without Premiere Pro or CapCut
- One step, no timeline. Drop the clip, pick the shape, download.
- Private. The video stays on your device; nothing is uploaded to an editing service.
- Sound untouched. The audio is copied; only the picture is re-encoded once.
- Free and unwatermarked, in any modern browser.
Tips & troubleshooting
Add captions after reframing, so they are laid out for the vertical frame: use the subtitle generator.
Long podcast? Let the AI clip maker pick the best moments first, then reframe only those.
Need an exact, fixed crop instead of tracking? Use crop video.
AI Auto Reframe (Vertical Video) FAQ
How do I auto reframe a video without Premiere Pro?
Drop the video here, choose 9:16, 4:5 or 1:1, keep Follow on Faces and press Reframe. The face-tracking crop is computed and rendered in your browser, with no editor, account or upload. The vertical video downloads as MP4, or as WebM or MOV if that was the source.
How do I convert horizontal video to vertical and keep the face in frame?
Use auto reframe with Follow: Faces. A face detector tracks the speaker and the 9:16 crop moves with them, smoothed so it feels like a camera operator rather than a jumpy auto-zoom. A plain centre crop often cuts off anyone who is not in the middle.
Is auto reframe free in CapCut?
CapCut’s plans and which features they include change over time, so check the app. This auto reframe is free with no account and no watermark, and it works in a browser on a computer or phone.
Does reframing lower the quality?
It crops, so a 1080p landscape video becomes a 608×1080 vertical one: fewer pixels, but each is as sharp as before. The picture is re-encoded once at the quality you choose, and the sound is copied unchanged. Start from 4K for a full-size vertical video.