How to convert a horizontal architecture video to vertical, and when to re-render instead

Your walkthrough was rendered at 16:9 for a client presentation or YouTube, and now it needs to run as a Reel, a TikTok or a Short. A vertical frame keeps less than a third of a widescreen image's width, so converting is never just a setting: you choose what to keep, what to lose, and whether the result still shows the building properly. This guide compares the three layouts (crop and pan, blur-fill, split-screen) with the pixel math for each, the auto-reframe tools in Premiere, DaVinci Resolve and CapCut, ffmpeg commands that do each layout, and the cases where re-rendering vertically is the better use of an evening.

Key takeaways

• A full-height 9:16 crop of a 1920 × 1080 video keeps only 608 of its 1,920 pixels of width, about 32%, and must be enlarged 1.78 times to fill a 1080 × 1920 frame, so it looks softer.
• From a 4K (3840 × 2160) master, the same crop is 1215 × 2160 and scales down to 1080 × 1920, so start from the highest-resolution export you have.
• Blur-fill keeps the whole widescreen composition but shrinks it to a band about a third of the screen's height; split-screen keeps 63% of the width of each of two shots.
• Premiere's Auto Reframe, Resolve Studio's Smart Reframe and CapCut's Auto reframe automate the crop, but they follow motion, so check that they follow the building rather than a passer-by.
• If you still have the model or the original renderings, a vertical render or a film made from the stills beats any reframe of a wide video.

How to convert a horizontal architecture video to vertical

To convert a horizontal architecture video to vertical, place it in a 1080 × 1920 (9:16) timeline and choose one of three layouts: crop to full height and pan the crop window across the building, keep the full widescreen frame over a blurred or solid background, or stack two widescreen shots as a split-screen. Auto Reframe in Premiere, Smart Reframe in DaVinci Resolve Studio and Auto reframe in CapCut can automate the crop. If the building is long and low, or you still have the 3D model, re-rendering vertically gives a better result.

Which layout works depends on the building and the shot, and the numbers below show why.

How much of the frame each method keeps

The geometry is fixed. A 9:16 frame cut from a 16:9 frame at full height is 9/16 × 9/16 of the width, roughly 31.6%. Everything else is a trade between keeping width and keeping size.

Two consequences matter for architecture. First, resolution: enlarging a 1080p crop by 1.78 times is visible on fine lines like mullions and cladding joints, so export or request the widescreen master at 4K if you can. Second, composition: a widescreen walkthrough was framed for width, and the crop will cut through whatever was placed off-center. Watch the whole video once with a 9:16 guide over it before deciding.

Crop and pan: when it works and how to keep it smooth

A full-height crop fills the phone screen, which is why it's the default on Reels and TikTok. It works when the subject is already vertical or central: a tower, a double-height living room, a stair, a doorway, a garden path walked straight down the middle. It fails on a long, low house, a street elevation or any shot whose point is the width.

When the subject moves across the frame, or the important part sits off-center, keyframe the crop window's position:

1. One move per shot. If the original camera already pans, keep the crop fixed and let the building slide through it. Adding a second move on top of the camera's own reads as seasickness.
2. Slow and steady. Ease in and out, and keep the speed constant between. A crop that darts to follow details looks automated.
3. Pan toward the destination. On a static wide shot, a slow pan from one end of the facade to the entrance turns a cropped shot into a reveal.
4. Never cut the building at a corner. If a crop must choose, keep one complete corner and lose the far end, not the middle.

Then check the safe zones: on a vertical platform the bottom third is covered by captions and buttons, so the building's base shouldn't sit there. The text and safe-zone guide gives the pixel values.

Blur-fill and solid backgrounds

Blur-fill places the full widescreen frame in the middle of the vertical canvas over an enlarged, blurred copy of itself. Nothing is cut, which is why it's the honest choice for a masterplan, a long elevation or a landscape where width is the subject. The cost is size: the image is a band about 608 pixels tall in a 1,920-pixel frame, and on a phone it looks like what it is, a horizontal video on a vertical platform.

Three ways to make it look intentional:

• Use a solid color instead of blur. A project color, or a neutral that matches the image, looks designed; a smeary blur looks like a repost.
• Use the empty bands. Put the project name above the image and a one-line description below it, inside the safe zone. The layout then reads like a card rather than a mistake.
• Enlarge the band slightly. Scaling the widescreen image to about 120% and cropping its sides a little gives a taller band and still keeps most of the composition.

Blur-fill is also the right choice for LinkedIn, where widescreen and square video sit comfortably in the feed; the LinkedIn video guide covers formats there.

Split-screen: two views at once

Stacking two shots, each 1080 × 960, keeps about 63% of each one's width, twice what a full crop keeps. It suits pairs that make sense together: before and after, exterior and interior, an aerial over a street view, a plan over the space it describes. Keep both halves moving slowly and in the same direction, give the eye one dominant half (usually the top), and add a thin divider line in a neutral color. A split-screen of two unrelated views is busy without telling the viewer anything.

Auto-reframe tools compared

All three tools analyze the footage and move a crop window for you. They're built to follow "the action," which in architecture video is often a person walking through the frame rather than the building, so treat their output as a first pass.

Sources: Adobe Help, Blackmagic Design, CapCut.

Adobe's notes are the most useful for architecture. Premiere "creates a duplicate sequence with the correct dimensions for the new aspect ratio" and applies the effect to every clip, so your horizontal edit stays intact. Of the three motion presets, "Slower Motion" is described as appropriate when "there is very little or no camera motion," with "almost static" results and few keyframes, which is closer to how architecture should move than "Default" or "Faster Motion." Adobe also warns that nesting "does not maintain soft transitions properly," and that complex sequences "may need to fine-tune the keyframes" afterwards. In every tool, scrub the result looking for two failures: the crop drifting after a person or car, and the crop cutting through the building's main corner.

Doing it with ffmpeg

If you'd rather not open an editor, ffmpeg (free, command line) does each layout in one command. These assume a 16:9 input called wide.mp4 and produce 1080 × 1920 H.264 files.

Center crop at full height:

ffmpeg -i wide.mp4 -vf "crop=ih*9/16:ih,scale=1080:1920:flags=lanczos,setsar=1" -c:v libx264 -crf 18 -preset slow -pix_fmt yuv420p -c:a copy vertical-crop.mp4

Crop placed a quarter of the way in from the left, for an off-center subject (change 0.25 to move it):

ffmpeg -i wide.mp4 -vf "crop=ih*9/16:ih:(iw-ow)*0.25:0,scale=1080:1920:flags=lanczos,setsar=1" -c:v libx264 -crf 18 -preset slow -pix_fmt yuv420p -c:a copy vertical-offset.mp4

Blur-fill:

ffmpeg -i wide.mp4 -filter_complex "[0:v]scale=1080:1920:force_original_aspect_ratio=increase,crop=1080:1920,boxblur=20:2[bg];[0:v]scale=1080:-2[fg];[bg][fg]overlay=(W-w)/2:(H-h)/2,setsar=1" -c:v libx264 -crf 18 -preset slow -pix_fmt yuv420p -c:a copy vertical-blur.mp4

Split-screen from two clips:

ffmpeg -i top.mp4 -i bottom.mp4 -filter_complex "[0:v]crop=ih*9/8:ih,scale=1080:960[a];[1:v]crop=ih*9/8:ih,scale=1080:960[b];[a][b]vstack,setsar=1" -map "0:a?" -c:v libx264 -crf 18 -preset slow -pix_fmt yuv420p -c:a copy -shortest vertical-split.mp4

ffmpeg can't follow a subject the way the auto-reframe tools do, but for a fixed crop it's precise and repeatable. If the upload then looks soft on Instagram, the blurry video guide covers the upload side.

When to re-render vertically instead

Reframing is damage control. A vertical original, composed for 9:16, is better whenever you can get one.

You still have the model. Real-time renderers can export vertical video directly, and a camera path designed for a tall frame looks deliberate in a way a crop never does: higher eye levels in tall rooms, more ground in the foreground, corners rather than flat elevations. The vertical export guide for Enscape, Lumion, Twinmotion and D5 gives per-program settings.

You only have the renderings. If the walkthrough came from a visualizer and you don't have the model, the still images behind it are often the better source. An Arch2Video film is generated from six images of one project and delivered as a 1080 × 1920 vertical MP4 of about 30 seconds, with six connected five-second AI-generated scenes. Landscape renderings can be used; each scene is framed vertically, so keep the important element of each image away from its far left and right edges. AI motion is an interpretation of the image rather than a measured walkthrough, so review it before posting, and label unbuilt work as concept work. The guide to turning renderings into video compares this with other routes.

The building is long and low. No crop rescues a 60-meter single-story elevation in 9:16. Either use blur-fill honestly, or show it through details and a walk along its length.

Frequently asked questions

How do I convert a horizontal video to vertical without losing quality?

Start from the highest-resolution master you have. A 9:16 crop from 1920 × 1080 keeps only 608 pixels of width and has to be enlarged 1.78 times, which softens fine lines, while the same crop from a 3840 × 2160 master is reduced instead and stays sharp. Export once at 1080 × 1920 with a generous bitrate, and avoid re-exporting the converted file through a second app.

Can I just upload a horizontal video to Reels or Shorts?

Instagram accepts Reels with an aspect ratio between 1.91:1 and 9:16, so a horizontal video uploads, but it won't fill the vertical screen and the building appears small. If width is the point of the shot, a deliberate blur-fill or solid-background layout with a title in the empty space looks better than an unmodified upload. For most project shots, a crop or a vertical re-render performs better.

Is Auto Reframe good for architecture footage?

It's a good first pass, not a finished edit. Auto-reframe tools follow motion, so in an architecture video they can drift after a person or car instead of holding the building. In Premiere, the "Slower Motion" preset, which Adobe describes as suited to little or no camera motion, usually suits slow architectural moves best. Check every keyframe for drift and for crops that cut through the building's main corner.

Should I reframe in CapCut, Premiere or DaVinci Resolve?

Use the tool you already edit in. CapCut is free and fastest for a single clip, though Auto reframe is only available in some regions. Premiere reframes a whole sequence into an editable duplicate, which helps with a long edited walkthrough. Smart Reframe needs the paid Studio version of DaVinci Resolve, worth it mainly if you already grade and finish there. For a fixed crop with no tracking, ffmpeg does the job for free.