How to Auto-Reframe Landscape Video to Vertical in 2026

Almost every long video is filmed wide. Podcasts, interviews, webinars, keynote talks, and gameplay are all captured in 16:9. Every short-form feed that matters - TikTok, YouTube Shorts, Instagram Reels - is vertical 9:16. That mismatch is the single most tedious step in turning long content into clips, because a naive crop either chops the speaker out of frame or leaves the subject drifting off to one side while empty background fills the middle. Reframing used to mean sitting in an editor and keyframing a crop window frame by frame so the camera follows whoever is talking. Auto-reframe does that work for you.

This guide explains how Shortzly's auto-reframe tool turns a landscape source into a clean vertical clip, the different layouts it can compose, how the Auto mode decides which one to use, and how to pick the right layout by hand when you already know what the footage looks like.

What Auto-Reframe Actually Does

A reframe is not just a resize. Shrinking 16:9 down to fit a 9:16 frame leaves black bars top and bottom, and letterboxed uploads get read by every short-form algorithm as low-effort repurposed content. The goal is to fill the vertical frame with the part of the picture that matters and to keep it there as the action moves.

Shortzly does this by analyzing the clip before it renders: it samples frames across the clip range, finds faces, measures how much the rest of the picture is moving, and checks how flat or busy different regions are. From that it builds a layout - a set of source rectangles mapped into the vertical output - and then renders the clip frame by frame so the crop tracks the subject with a smooth, eased camera rather than a jittery lock. The result looks like the clip was shot vertically in the first place.

The Layouts

Not every clip wants the same treatment. A single talking head, two people in conversation, and a streamer with gameplay behind them each need a different composition. Shortzly supports six:

• Auto. Shortzly inspects the footage and chooses the layout that fits. This is the default, and for most mixed source material it picks the right one without you thinking about it.
• Speaker focus. A single tracked crop that follows the active face around the frame. This is the workhorse layout for one-person content - a solo creator, a presenter, a single talking head in a wide shot.
• Split-screen. For two or three people sitting too far apart for one crop to hold them all, Shortzly gives each person their own panel. On a tall 9:16 output the panels stack vertically; on a wide output they sit side by side.
• Webcam over gameplay. The gaming layout keeps the streamer's webcam and the gameplay both in frame, so a clip reads as a reaction and the action at the same time. This is the layout behind the gaming clip maker.
• Screen plus speaker. For tutorials and demos where a screen recording fills most of the frame and the presenter sits in a corner, Shortzly keeps the screen readable and the speaker visible together.
• Center. A fixed center crop with no tracking. It is the fastest option and the right call when the subject is already centered and steady, or when you deliberately want a locked frame.

How Auto Mode Decides

The interesting part of Auto is that it reads the structure of the footage rather than guessing from the file. A small, steady face near the edge of the frame is almost certainly a webcam. From there, Shortzly looks at what the rest of the picture is doing. If the background is moving - the hallmark of live gameplay - it chooses the webcam-over-gameplay layout. If the rest of the frame is flat and static in the way a screen capture is, it chooses screen plus speaker. If it is neither, the person is simply in a wide shot and gets speaker focus.

When there is no obvious webcam, the decision comes down to how many steady faces there are and how far apart they sit. Two or three people spread across a wide frame get split-screen, because one crop cannot hold them without shrinking everyone. A single person gets speaker focus. And when there is no usable face at all, Shortzly falls back to a center crop rather than inventing a subject to track. If you request a layout the clip cannot actually fill, it degrades gracefully to speaker focus, or to a plain center crop when no face is present, instead of producing a broken frame.

Face Tracking Under the Hood

Speaker focus and any tracked panel rely on face tracking, and Shortzly offers three modes:

• OpenCV is the fast default. It locks onto the first confident face and follows it with a sub-pixel crop and a small dead zone, so the camera does not twitch on every tiny movement and the clip does not open with a distracting slide in from center.
• MediaPipe is the accurate option, roughly two to three times slower. It uses a face mesh with lip-activity scoring, which is what lets it tell who is actually speaking and switch focus between people in a conversation.
• None skips detection entirely and center-crops. It is the fastest path and pairs naturally with the Center layout.

One useful detail: when you set a layout explicitly, the layout decides the framing. Face tracking mode only matters for choosing MediaPipe on speaker focus. If you want a fixed, untracked crop, the Center layout is how you ask for it.

Every Clip, Every Aspect Ratio

Vertical 9:16 is the default because it is what TikTok, Shorts, and Reels want, but Shortzly renders the same reframe into 1:1 for square feeds, 4:5 for the taller Instagram feed slot, and 16:9 when you need the original wide shape back for a YouTube re-upload or a landscape player. You can render several ratios from a single job, so one reframe decision covers every platform at once. The one case where no reframing happens is 16:9 to 16:9, where the source is already the target shape and Shortzly simply resizes.

How to Reframe a Clip in Shortzly

1. Bring in a long video. Paste a YouTube, Vimeo, or Twitch URL, or upload a file. The AI clip generator finds the moments worth clipping first, so you are reframing highlights rather than the whole recording.
2. Pick a highlight to render. Open any moment and choose your output aspect ratio - 9:16 for vertical feeds is the usual choice.
3. Leave the layout on Auto, or set it yourself. Auto fits most footage. If you already know the clip is a two-person interview or a streamer with gameplay, set split-screen or webcam over gameplay directly.
4. Choose a face tracking mode if it matters. OpenCV is fine for a static solo setup. Switch to MediaPipe when focus needs to move between speakers.
5. Render. Shortzly composes the layout, tracks the subject frame by frame, and burns animated captions on top. The output drops straight into your review queue, ready to publish.

Picking the Right Layout for Each Platform

The platform rarely changes which layout is correct - the footage does - but a few habits help. For TikTok and Reels, where the frame is small and the first second decides everything, speaker focus on a single face reads cleanest and keeps the subject large. For interview and podcast content headed to any vertical feed, split-screen keeps both voices visually present, which matters when the conversation is the content. For tutorials going to YouTube Shorts, screen plus speaker keeps the demo legible while the presenter stays on screen, which is exactly what a how-to audience is there for.

Whatever the layout, the vertical crop should put the subject in the upper two-thirds of the frame so platform captions and UI at the bottom do not cover faces. Shortzly's framing already biases toward this, which is part of why an auto-reframed clip tends to sit better in a feed than a hand-cropped one that was centered for a landscape player.

Key Takeaways

• Auto-reframe fills a vertical 9:16 frame with the subject and keeps it there as the action moves, instead of letterboxing a wide source.
• Six layouts cover the real cases: Auto, speaker focus, split-screen, webcam over gameplay, screen plus speaker, and a fixed center crop.
• Auto mode reads the footage - a steady edge face is a webcam, a moving background is gameplay, a flat background is a screen capture - and chooses accordingly, falling back safely when a clip cannot fill a requested layout.
• OpenCV tracks fast, MediaPipe follows the active speaker, and None center-crops; when a layout is set, the layout decides the framing.
• Render 9:16, 1:1, 4:5, and 16:9 from one job so a single reframe serves TikTok, Shorts, Reels, and every other placement.

Want to see it on your own footage? Start with a free Shortzly account, paste a wide interview or stream, and watch a landscape source become a vertical clip that tracks the speaker without a single keyframe.