Convert horizontal video to vertical on Mac (AI reframe, 2026)

You need to convert horizontal video to vertical on a Mac, and the footage is already sitting there: a 16:9 podcast recording, an interview shot on a tripod, a screen-share webinar, a talk filmed from the back of the room. The platforms you’re posting to — TikTok, Reels, YouTube Shorts — want 9:16. And the obvious move, cropping the center of the frame, works for exactly one kind of shot: the one where the subject sits dead center and never moves.

Everything else breaks. Two people talking on a podcast set means one of them is always outside the center crop. A speaker who paces slides out of frame and back in. An off-center tripod setup gives you thirty minutes of a shoulder. Converting horizontal to vertical isn’t really a resizing problem — it’s a tracking problem, and the tools that treat it as resizing produce clips that look like security-camera footage cropped by accident.

This post covers what a good horizontal-to-vertical conversion actually requires, the three ways to do it on a Mac — manual keyframing, cloud auto-reframe, and native on-device AI — and where each one earns its place. The short version: if you’re converting long-form footage into clips regularly, the native route is faster, free, and doesn’t involve uploading a 6 GB recording to anyone’s server.

Why the center crop fails: the arithmetic

Start with the numbers, because they explain every bad vertical crop you’ve seen.

A 16:9 source at 1920×1080 converted to 9:16 at 1080×1920 keeps a window just 608 pixels wide out of the source’s 1920. That’s 31% of the horizontal frame. You are discarding more than two-thirds of every frame, and the entire quality of the output depends on which two-thirds.

On a solo, centered, locked-off shot, the center 31% contains the face. Fine. On a two-person podcast frame, the center 31% contains the table between the hosts. On a conference talk, it contains the slide edge and half a lectern. The information you’re posting the clip for — the face of whoever is talking — lives in the left or right third, and a static crop can’t reach it.

So the real job is a moving window: a 9:16 region that sits on the active speaker’s face, hands off to the other speaker when the conversation turns, and moves smoothly enough that the motion reads as intentional camera work rather than jitter. That’s what “auto-reframe” means, and the quality bar is whether you’d believe a camera operator did it.

Option 1: manual keyframing in Final Cut or Premiere

The traditional answer on a Mac is to do the tracking yourself. Final Cut Pro and Premiere Pro both handle it the same way in principle: drop the 16:9 clip into a 1080×1920 vertical timeline, scale it up so the frame height fills the canvas, then keyframe the horizontal position so the visible window follows the subject.

It works, and for a single hero clip it’s the maximum-control option — you decide every reframe, every hold, every cut between speakers. But be honest about the cost:

It’s per-clip labor. A 60-second two-person clip needs a position keyframe at every speaker change, each one placed and eased by hand. Ten to twenty minutes per clip is a realistic estimate once you include review passes.
It doesn’t scale. If your week’s output is five to ten clips from a two-hour recording, manual reframing alone is a half-day of editing — before captions, before selection, before export.
Final Cut costs $299.99 and Premiere is $22.99/month — reasonable if you’re already an editor living in those apps, absurd as a purchase just to crop video vertically.

Manual keyframing is the right tool when one specific clip matters enough to art-direct. It is the wrong default for volume.

Option 2: cloud auto-reframe tools

The web tools — Opus Clips, Vizard, Klap, and the rest — bundle auto-reframe into their clipping pipelines, and the reframe itself is generally competent. The problems are everything around it:

The upload. This is the Mac-specific pain. Long-form source files are big — a two-hour 4K podcast recording is 10–20 GB. On a typical home upload connection that’s an hour or more of progress bar before processing even starts, per episode, forever. Your M-series Mac, which could chew through that file locally in minutes, sits idle while the footage crawls to someone else’s data center.
The meter. Cloud tools price by upload minutes — typically 100–300 minutes a month on the $15–30 plans. A weekly two-hour show eats that allowance fast, and a busy month means overage or a plan bump.
The subscription. $180–360 a year, every year, to run a workload your own hardware handles.
The privacy default. Every full recording — including the pre-roll chatter, the parts you’d never publish — lands on a third-party server under a terms-of-service you skimmed.

If you’re on Windows or need built-in social scheduling, the cloud tools are a defensible choice. On an Apple Silicon Mac, you’re paying rent to avoid using the computer you already bought.

Option 3: on-device AI reframe with Clipolette

Clipolette is a native app for Mac (Apple Silicon, macOS 15.4 or later), iPad, and iPhone that turns long-form video into short vertical clips, and the entire pipeline runs on-device: moment selection, transcription, captioning, and the auto-reframe. Converting horizontal to vertical isn’t a separate feature you configure — it’s what happens when you pick a 9:16 output for 16:9 footage.

The workflow on a Mac:

Drop the source file in. MP4 or MOV, straight from the camera, OBS, Zoom, or a downloaded livestream. Nothing uploads; the 15 GB recording never leaves the SSD.
Choose 9:16 (or 1:1 if you’re posting square). This is the conversion decision — everything downstream frames for it.
Let the AI pick the moments, or steer it. The selection takes plain-English direction — “pull the moments where the guest tells a complete story” — so you’re converting the parts worth posting, not the whole timeline. (If you want the full timeline converted, that works too.)
The reframe runs during processing. The crop window tracks faces and action per-frame on the Neural Engine, holds steady when a speaker holds still, and hands off between speakers instead of drifting across the set.
Review, fix the transcript, export. Captions burn in after you’ve had a pass at the transcript, and the batch lands in a folder as platform-ready vertical files.

On an M-series Mac the pipeline runs through an hour of footage in minutes, and there’s no meter — the whole backlog can go through in one sitting, the same way the batch export workflow handles it. The app is free with every feature unmetered; a one-time $24.99 purchase removes the export watermark, and that’s the entire price list. No subscription.

The same reframe pipeline runs on iPhone — covered in the vertical video cropping AI on iOS guide — but the Mac has two practical advantages for this specific job: the big screen makes crop review honest (a bad reframe hides on a 6-inch display and glares on a 27-inch one), and the source files are usually already on the Mac, because that’s where the recording, the podcast session, or the screen capture happened.

What to check before you trust any auto-reframe

Whatever tool you use, the failure modes are the same. Run one representative clip through and look for these before you commit a whole backlog:

Speaker handoffs. The hardest case is two people alternating quickly. A good reframe cuts or pans decisively at the turn; a bad one drifts to the midpoint and frames the gap between two half-faces. Test with the fastest back-and-forth section you have.

Stability on a static subject. When nobody moves, the window should lock. Constant micro-adjustments — the crop “breathing” around a stationary face — is the tell of a naive per-frame tracker, and it’s nauseating at fullscreen phone size.

Off-center sources. Feed it a shot where the subject sits in the left third. Center-crop tools fail this instantly; real trackers don’t care where the subject starts.

Graphics and screen shares. If the footage cuts to slides or a screen share, watch what the crop does. Faces are trackable; a 16:9 slide crushed into 9:16 is unreadable no matter how good the tracking is, and the honest answer for slide-heavy footage is choosing moments where the speaker carries the point — which is a selection problem before it’s a cropping problem.

Your own review pass. No tracker is right 100% of the time. The workflow question is whether you can see the reframe before export and reject a clip that framed wrong — cheap on a Mac screen, annoying through a cloud tool’s preview player.

Making the source footage easier to convert

If you control the recording setup, three habits make every future conversion better:

Shoot in 4K if the camera has it. The vertical window from a 4K source is 1215 pixels wide — wider than a full-HD vertical output needs — so the crop involves no upscaling and the output stays sharp. From a 1080p source, the 608-pixel window gets scaled up to 1080 wide, which is acceptable but visibly softer.

Frame subjects a third in from the edges, waist-up. A medium shot gives the vertical window a subject that fills the frame. A wide shot of the whole set means even a perfectly tracked crop shows a small person in a big room.

Keep the speakers apart in the frame. If two people sit nearly shoulder to shoulder, no 9:16 window can isolate one cleanly. A bit of separation lets the reframe cut between clean singles — which is exactly how the interview-to-Reel workflow gets clips that look shot vertical.

None of this is new gear — it’s the same framing discipline that makes footage easier to edit generally, and it compounds with every clip you cut.

The bottom line

Converting horizontal video to vertical on a Mac is a tracking problem wearing a resizing costume. Manual keyframing in Final Cut or Premiere solves it with your time — worth it for one hero clip, brutal at volume. Cloud auto-reframe solves it with your money and your upload bandwidth — $180+ a year plus an hour of uploading per long recording. A native on-device pipeline solves it with the Apple Silicon you already own: Clipolette reframes, captions, and exports vertical clips from 16:9 footage in minutes, locally, for free, with a single one-time $24.99 watermark unlock as the only price.

The test is one file: get Clipolette on the App Store, drop in your last horizontal recording, pick 9:16, and compare the tracked reframe against the center crop you were about to settle for. It also slots into the broader long-form to short-form Mac workflow if clips are a weekly habit rather than a one-off.

FAQ

How do I convert horizontal video to vertical on a Mac for free? Clipolette runs a full AI auto-reframe on-device on any Apple Silicon Mac: drop in the 16:9 file, choose 9:16, and the crop window tracks faces through the footage automatically. The app is free with no minute caps; a one-time $24.99 purchase removes the export watermark. The manual alternative — keyframing position in iMovie or Final Cut — is free-to-cheap but takes 10–20 minutes per clip.

Does converting 16:9 to 9:16 lose quality? You keep about a third of the horizontal frame, so resolution matters: from a 4K source the vertical window is wider than 1080 pixels and exports sharp; from a 1080p source the window is 608 pixels wide and gets upscaled, which looks acceptable but softer. Shoot 4K when you know vertical clips are the destination.

Can AI reframing handle two people talking? Good trackers handle it by cutting or panning between speakers as the conversation turns — that’s the case to test before trusting any tool. Feed it your fastest back-and-forth segment: if the crop drifts to the midpoint and frames the space between two half-faces, the tracker is naive and will do it on every clip.