How to Convert Horizontal Video to Vertical Without Losing the Speaker

The output is the same clip, just vertical: 9:16, square 1:1, or 4:5. It's the whole thing from start to finish — not a highlight cut — where the frame leads the speaker. That fits an interview, a podcast, or a talk: the full conversation as a vertical for Reels and Shorts, square and 4:5 for the feed where you don't need a full-height frame.

The difference from a plain crop is that the window doesn't stay put. It moves to whoever's talking and pushes in a little on the face. In a two-person conversation that's what kills the empty center: when the guest on the left answers, the frame slides left; when the right one jumps in, it slides right.

In a landscape interview the guests sit near the edges, and the center of the frame is a table, a mic, or a blank wall. Cut a vertical slice out of the middle and you capture exactly that empty space, with the speaker half-buried behind the frame edge. A static crop in an editor holds one position for the whole clip — it has no idea who's talking and can't reposition itself.

Speaker-tracking reframe works differently: it reads who's speaking and the moving crop window follows them. On turn changes the window hops from one guest to the other, the way an editor would cut between shots by hand. Check it yourself: take 20 seconds of overlapping dialogue and watch whether the speaker's face stays centered in the vertical frame.

In Kutto this is Auto Reframe under AI tools. It takes the horizontal clip whole and hands back the same clip in the vertical ratio you pick — no manual keyframing of shots.

1. 1

   Give it a horizontal clip

   Upload a file, pick a clip from your library, or paste a link. You don't need to work out the length up front — a link with no known length takes a top estimate, and you're billed by what it actually runs.

2. 2

   Pick the output ratio

   9:16, 1:1, or 4:5. The ratio sets the crop width and how tightly the frame pushes in on the face, so it's not just a canvas size — it decides how close the camera sits to the speaker.

3. 3

   Wait for analysis and assembly

   Auto Reframe first passes the clip in windows, building a timeline of who's talking and where the camera moves, then assembles the vertical in pieces and attaches the source audio. A long clip builds gradually, with progress on the job row.

4. 4

   Grab the finished clip

   The result lands in your library as a new asset. From there you can download it or hand it straight to captions or clipping without exporting to your computer first.

9:16 fills the whole screen — that's the format for Reels, Shorts, TikTok, and any full-screen feed. But in a scrolling feed you don't always want full-height video: 4:5 and square often work better because they take less vertical space and leave room for the caption. The ratio also changes how tightly the frame sits on the speaker's face, not just the shape of the canvas.

If you're not sure, start with 9:16 — it's the most common ask for Shorts and Reels. Make a square or 4:5 separately only if the same clip is also going out as an in-feed post.

Auto Reframe is an AI tool, so it needs a paid plan — the free Start plan has no AI at all. On a paid plan it bills source minutes from your quota, the same way clipping does, and in your history it shows up as its own "Auto Reframe" line. Minutes are reserved when the job starts; a link with no known length takes a top estimate, and the final count is set by a probe of the file. Overage past your quota only runs with your consent, and if the job fails or you cancel it, the minutes come back.

Processing isn't instant: the clip is analyzed in five-minute windows, so a one-hour source builds up gradually rather than in a single pass. Analysis and assembly also survive a worker restart — if the process drops partway through a long clip, it resumes from the piece where it stopped instead of starting over. Current plan pricing is on the pricing page.

Auto reframe steers the camera by who's talking. Where there's clearly no one speaking, or the speaker changes too fast, the window can land in the wrong spot. That's normal for any automatic framing — the point is knowing where to look and patching the rough stretch by hand.

• Both silent. Pauses, music breaks, an intro card — there's no one to follow, so the frame may sit on the wrong guest.
• Fast cross-talk. When lines overlap and speakers swap inside a second, the window can't reposition cleanly and you'll see extra seams on the cuts.
• People at the edge of the 16:9 frame. If a guest sits right against the edge, pushing in can clip part of their face past the vertical window.

On a rough stretch, try a different ratio first — 1:1 and 4:5 have a wider window, so the speaker falls off the edge less often. If the problem is a short segment, re-run just that piece as its own job and drop it into your edit. And for dialogue where you need both people on screen at once, Split-screen or building the clip in clipping can work better, since the frame follows a plan rather than one speaker.

Before you post, run the finished clip through a short checklist — pay the most attention to the turn changes, where the frame repositions.

• Speaker in frame on turn changes, face not clipped by the edge
• Audio present and in sync, last frame not cut off
• Captions, if you added them, keep up with the speech and fit inside the vertical frame
• Ratio matches the platform: 9:16 for Shorts and Reels, 4:5 or 1:1 for the feed
• File length and size within the target platform's limits

Pick one horizontal interview or podcast where the guests sit near the edges — that's where the difference from a center crop shows up right away. Run it through Auto Reframe in 9:16, watch the turn changes and the spots where someone goes quiet, then fix the handful of places that drifted.