How to turn one finished video into Reels, Shorts and feed versions without cutting off the subject

One finished video often needs to live in four places: Instagram Reels, YouTube Shorts (YouTube's short vertical video format), the Instagram or Facebook feed, and sometimes a wide version for a website. Each place shows a different shape. If you simply stretch or crop, the speaker leaves the frame, or your logo ends up under the platform buttons.

This page is only about a video that is already finished. It covers the order I would work in: free tools first, then Higgsfield, an online AI image and video platform, only for the one case where the free tools cannot do the job. For photos, see my page on changing the shape of an image.

This page may contain affiliate links from which Blur Studio earns commissions.

Which direction are you resizing?

Decide this first, because the two directions are very different jobs.

• Shrinking the shape: wide 16:9 to tall 9:16. You throw pixels away. A tall frame cut from a wide one keeps only about a third of the width, so you lose roughly 68% of the picture sideways. The whole job is choosing which third to keep, scene by scene. No AI is needed. A tool that follows a face does most of it for you.
• Growing the shape: tall 9:16 to wide 16:9. You need pixels that were never filmed. Options are a blurred or branded background behind the tall video (nothing invented), or an AI tool that invents the missing sides. Invented sides can look wrong, so you must watch the result.

If you can still choose how to film, film wide and keep the subject in the centre third. Then every crop works. This one habit saves more time than any tool.

The sizes you actually need

Aspect ratio is the shape of the video, width to height. Most businesses need three shapes from one master.

Ratio	Where it is used	Size to export
9:16 (tall)	Instagram Reels, YouTube Shorts, TikTok, Stories	Meta's ad spec for Reels is 1440 by 2560. Many editors and vendor guides use 1080 by 1920 for organic posts, which is a common working size, not a Meta rule
4:5 (slightly tall)	Instagram and Facebook feed	Meta's ad spec for Feed video is 4:5, 1440 by 1800, MP4 or MOV, up to 4 GB
1:1 (square)	Older feed posts, some marketplaces, carousels	Check the platform page for the exact size you need
16:9 (wide)	Website, YouTube, presentations	Usually your master

The Meta figures above are advertising specs from Meta's own pages, as of November 2026. Meta's aspect ratio help page lists 4:5 and 1:1 for Feed and 9:16 for Stories and Reels, and tells advertisers to review auto-generated crops so nothing important is cut off. Meta's Reels ad spec allows files up to 15 minutes and 4 GB. Facebook Reels specs can differ, so check them if you post there. Platform rules change, so confirm on the platform's current help page before a big campaign.

Step 1: keep a clean master with no burned-in text

Burned-in text means words and logos that are part of the video file, not a separate layer. When you reframe a finished video, burned-in captions move and get cut along with the picture. A caption that sat nicely at the bottom of a wide frame may be half outside the tall frame.

So before you resize anything:

1. Export one master with no captions, no lower-third name bars and no logo. Name it clearly, for example clientname_launch_MASTER_nocaps.mp4.
2. Keep the caption file (an SRT file is a plain text file of words and times) and the logo file separate.
3. Make each shape from the clean master, and add captions and logo last, placed for that shape.

If you only have a finished video with captions already burned in, a re-edit from the original project beats any reframe. If the original is gone, plan to crop tighter and check that the burned-in words survive in every scene.

Step 2: shrink wide to tall with subject tracking (free tools first)

Auto reframe is a feature that follows a person or object and moves the crop for you. Two free-to-start options, both from vendor pages read in November 2026:

• CapCut, a video editing app, lists Auto reframe on its desktop version. The steps on its page are: open the Video panel, turn on Auto reframe, choose the ratio, set the stabilization and camera speed, then apply. CapCut says it is available in specific regions, so it may not appear for you. It also tells you to preview that the subject stays in frame.
• Adobe Express, Adobe's online design and video tool, has a free video resizer with presets for Instagram, TikTok, YouTube and Facebook. Its page says uploads can be up to one hour. The page gives two different file size limits (1 GB and 2 GB), so check the limit on screen. It can also zoom and recenter the subject after the resize.

Adobe Premiere Pro, Adobe's professional video editor, also has an Auto Reframe feature. If you already own Premiere, read Adobe's help page for it to see how it works and which versions have it.

Where free auto reframe works well: a single talking person, a presenter, a podcast clip. Where it fails: a product close-up, a screen recording, two people on opposite sides of the frame, or a wide landscape shot where the subject is a tiny thing. In those scenes the tool follows a face that is not the point. Set the crop position by hand with keyframes (points on the timeline where you set where the crop sits). It takes a few minutes per scene.

A cheap option for some videos: keep the wide video whole in the middle of the tall frame and fill the top and bottom with a blurred, enlarged copy of the same video, or a brand colour. Nothing is invented and nothing is cut off. Many editors do this in one click.

Step 3: when the sides must be invented (Higgsfield Reframe and others)

Sometimes you have a tall video and need a wide one, or a square video that must become tall without cutting anything out. Neither a crop nor a blurred background looks right. This is the case for an AI extend tool.

Higgsfield has a tool called Reframe. Higgsfield's Reframe page says it changes a video's aspect ratio, follows the subject, and generates new picture where the original frame does not reach, instead of stretching or cropping. It lists the ratios 9:16, 16:9, 1:1, 3:4, 4:3 and 21:9 (the page is not consistent about 4:5, so check the ratio you need is offered). You can optionally add up to two reference images so the invented area matches your footage. Higgsfield's tool list gives output from 480p to 1080p. The same page contradicts itself on length: one place says clips up to 1 minute, another says any length, so test your clip length. Reframe is also a plugin for Adobe Premiere Pro (Higgsfield's plugin page says version 2025 and newer; another Higgsfield page says 2024 and newer, so check before installing). Higgsfield does not publish what Reframe costs in credits (credits are the points you spend on each generation), so check the price on screen before you confirm. For how credits behave in general, see how to spend Higgsfield credits wisely, and for the other edit tools see my guide to Higgsfield upscale, relight and edit tools.

What to expect, and the limits:

• Higgsfield says Reframe invents the missing area. It does not promise the invented area will look right, so judge it on your clip and do not assume it will be clean. Quality will vary by clip.
• The listed 1080p top output is lower than Meta's recommended 1440 by 2560 for Reels ads. For organic posts that is usually fine. For paid ads, check the platform's current requirement.
• Higgsfield's Reframe page says exported videos have no watermark, while Higgsfield's pricing page marks free output as watermarked and commercial use as not included on Free (its help center says Terms do not tie commercial use to a plan). That is a conflict, so check what you get on screen, and for client work use a paid plan. See the pricing explainer and higgsfield.ai/pricing.
• Faces, hands, text on signs and moving products are the usual places where invented sides go wrong. Watch the edges of the frame, not the middle.

Test it on one clip before you spend credits on twenty. The plan is below.

Step 4: safe zones using Meta's numbers

A safe zone is the part of the frame that stays clear of buttons, captions and the username. Keep your key words and logo inside it.

The only figures I would trust come from Meta itself. For Instagram Reels ads, Meta's specs list safe zones of 14% at the top, 35% at the bottom and 6% on each side. Read those as ad numbers. Meta does not say that organic Reels use the same margins, so for organic posts treat them as a careful starting point.

I would ignore pixel tables from blogs that give no source. One of them contradicted itself, and a vendor guide applied ad numbers to organic posts as if they were the same thing. For TikTok, YouTube Shorts and Facebook Reels, the official margins are not confirmed here, so look at each platform's preview in its own app before you rely on any number.

How to use it: on a 9:16 frame, shade the top 14% and bottom 35%. Keep faces, captions, price and the call to action (the line telling people what to do next) in the middle band. Save that overlay as a template in your editor.

Step 5: rebuild captions for each shape

Do not reframe text. Re-place it. For each version:

1. Import the caption file into the new, resized project.
2. Set the font size so a caption line holds about 4 to 6 words. A tall frame is narrower, so lines wrap more.
3. Move the caption block into the safe zone, above the bottom 35% on Reels.
4. Re-place the logo so it is not under the platform buttons.

It is faster to do this once as a template and reuse it across every client video. A weekly series becomes mechanical. See short-form video ads for clients for a full production checklist, and resizing ads for every placement if you are working from a designed ad rather than footage.

The phone check with pass and fail

Never judge on a computer screen. Send each export to your own phone and view it inside the app (or in a draft post that only you can see). Then score it.

Check	Pass	Fail
Subject in frame in every scene	Face or product fully visible in all scenes	Head or product cut off in any scene
Captions	All words visible above the bottom buttons	Any word hidden or cut at the edge
Call to action	Visible for the last 2 seconds, clear of buttons	Under the caption block or profile name
Logo	Not covered, not touching the edge	Covered or cropped
Invented areas (AI only)	No warped faces, hands or text at the edges	Any warping or odd shapes
Grid crop	Cover frame still reads as a 4:5 or square crop	Key words lost when the profile grid crops the cover

One test I would run: take a 30 second clip, export 9:16, 4:5 and 1:1, and score all three. If two shapes pass and one fails, fix that shape by hand rather than starting over.

Export and naming

Name every file with the client, clip, version and ratio so the wrong shape never gets posted. A pattern that works: client_clip_v3_9x16.mp4, client_clip_v3_4x5.mp4, client_clip_v3_1x1.mp4. Keep the master and the caption file in the same folder. For where to store finished files and prompts so nothing gets lost, see where to keep AI images, videos and prompts.

Platforms recompress uploads, so export at a good quality and do not export a tiny file to save space. If you are posting the same clip to several apps, upload each platform's own export instead of downloading it from one app and uploading to another, which adds a second round of compression and often a watermark.

Who should skip an AI reframe

• You only shrink wide to tall on a talking-head clip. Free auto reframe in CapCut or Adobe Express is enough. Do not spend credits.
• Your video has burned-in captions or on-screen text you cannot remove. Re-edit from the original instead.
• Your video is a screen recording or a product demo. Hand-set crop keyframes or a redesigned layout beat any automatic tool.
• It is a paid ad with strict spec rules. Use the editor you trust and check the platform's current specs.
• You need a result you can promise a client without trying. No AI reframe can promise that. Test first.

Higgsfield Reframe is a weak fit for most of these. It is worth a try when you have a finished tall or square video that must become a different shape and a blurred background looks cheap.

How I'd approach this

I would build the master once, with no burned-in text, and film wide with the subject in the middle. I would use a free auto reframe tool for the tall version, check it on my phone against the table above, and re-place captions inside Meta's safe zone numbers. I would only reach for an AI tool such as Higgsfield Reframe for the one version where sides really have to be invented, and I would try it on a single clip first.

If you make many videos a week for clients, the part worth automating is the routine: a master folder, a naming rule, a caption template per shape and a sign-off step. Blur Studio builds repeatable workflows like that, including approval portals and shared libraries around these tools, and you can see how that works.