3 best AI video expanders: outpainting, limits & exports

Runbo Li
Runbo Li
·
· 11 min read
Editorial illustration of a portrait coastal-road frame expanded into a landscape canvas; conceptual artwork, not a generated product result.

Quick answer

For expanding a video frame without cropping, start with Magic Hour for a short browser trial, Luma Reframe for canvas and prompt controls, or VideoExpander for a dedicated outpainting workflow. Choose Premiere Auto Reframe when cropping is acceptable and you need the subject tracked inside a new frame. Expansion adds generated surroundings; it does not make a clip longer or prove that its original pixels stay unchanged.

How we selected these tools

Magic Hour publishes this comparison and includes its own product. We selected three services whose first-party pages document video outpainting, plus one clearly labeled cropping alternative. Product documentation and public interfaces were reviewed October 13, 2026. This is a workflow comparison, not a matched output-quality benchmark. We did not generate the same clip in every product or measure a universal realism, speed or cost winner.

The decision is narrower than choosing an AI video generator. You already have a clip worth keeping, but its canvas does not fit the destination. A vertical interview may need a widescreen website version; a landscape product demonstration may need a vertical ad. The useful question is whether the whole composition must remain visible, which surrounding areas can safely be invented, and how much review the expanded file needs.

Tool

Best fit

What to verify

Magic Hour

Short browser trial and connected media workflow

Guest5second limit versus signed-in quote; watermark; accepted export

Luma Reframe

Canvas placement and prompt-guided surroundings

Selected version; conflicting12/16second documentation; actual audio

VideoExpander

Dedicated video outpainting controls

Free10second boundary; finished-file rights and settings

Premiere Auto Reframe (cropping alternative)

Track a subject when a crop is acceptable

What the new crop removes; tracking around cuts

Expansion, cropping, extension and upscaling solve different problems

Expansion, also called video outpainting, generates visual content beyond the edges of an existing frame. Cropping discards part of the view to fit another shape. Auto reframing moves that crop to follow a subject. Padding puts the original inside a larger canvas with plain or blurred borders. Stretching changes proportions. None of those operations should be presented as equivalent.

Extension adds time: a shot becomes longer. Upscaling increases pixel dimensions: the frame has more pixels but does not necessarily show more of the scene. When a brief says 'make the video bigger,' ask whether it means a different aspect ratio, higher resolution or a longer duration before choosing a tool. Otherwise, a technically successful generation can still solve the wrong problem.

Generated surroundings are plausible additions, not recovered evidence of what was outside the camera. Avoid outpainting documentary, insurance, forensic or other evidentiary footage when viewers could mistake invented content for the original recording. For an ad concept or stylized background, review the added area as a new creative asset.

1. Magic Hour: best for a short browser expansion trial

Magic Hour AI Video Expander accepts an uploaded clip and a destination aspect ratio. Its public page currently describes three free guest expansions per day, up to five seconds each, without signup. Free video results may include a watermark. Those trial rules do not establish the length, resolution or price of a signed-in project.

The Dashboard guide describes Add Video, a selected source segment, Aspect Ratio, the displayed credit estimate and Expand Video. It documents a trimmed segment up to 30 seconds and custom ratios between 1:8 and 8:1. Check the current validation and quote for your selection before submitting; a long uploaded file does not authorize processing its entire length.

Choose this route when a short existing clip needs a new publishing shape and you want a direct path into other media tools. Use the guest page to assess a representative shot, then inspect the whole result before buying more attempts. For a client delivery, separately confirm current commercial terms, source rights, watermark treatment and the final file requirements.

The practical boundary is review. An expansion can add an unwanted object, change a reflection or create an unstable background. 'Keep the original' is an acceptance requirement to verify, not a guarantee that every original pixel is unchanged. Compare the face, product, text, geometry and timing against the source as well as inspecting the new edges.

First-party sources: Magic Hour AI Video Expander · Dashboard guide

Magic Hour public AI Video Expander interface with video upload, expansion button and three remaining guest attempts, captured October13,2026.

Public Magic Hour interface captured October13,2026. This screenshot shows controls and an allowance badge; no output test is claimed.

Try one representative shot

Choose a short clip whose subject and original framing you need to keep. Review the new edges and exported file before a larger project.

Try AI Video Expander

2. Luma Reframe: best for positioning and prompt-guided surroundings

Luma’s Reframe guide documents dragging and scaling an image or video on a canvas, adding a prompt and choosing a target aspect ratio. It also describes a custom start frame and a transparent canvas download in the web workflow. Choose it when positioning the source and directing the surrounding scene matter.

Luma’s current Ray3.2 FAQ says Reframe preserves original audio. That differs from its older Reframe guide, which describes visual-only output and older upload constraints. The current page also gives inconsistent source-length figures: an overview says up to 16 seconds, while a specific Reframe answer says up to 12. Verify the selected interface’s limit rather than assuming every Luma workflow shares one duration or audio rule.

Save the original audio separately before adopting an audio-less output. Align it to the exported video and check duration, frame rate, spoken timing and the first and last frames. Do not assume that a visually similar expanded shot is automatically ready to replace the original on an edited timeline.

First-party sources: Current Ray3.2 FAQ · Older Reframe guide

3. VideoExpander: best for a dedicated outpainting interface

VideoExpander’s first-party outpainting page shows target-ratio, position, padding, quality and outpaint-pass controls. It accepts MP4, MOV or WebM. The public interface states that free expansion covers the first ten seconds; longer and advanced workflows require Pro. Check the resulting export and current account conditions rather than treating 'free' as unlimited finished footage.

The page also exposes multi-shot and shot-revision controls. That makes it worth evaluating when outpainting is the main job rather than one feature inside a larger generation platform. The controls shown in a public interface establish a documented workflow, not a quality or throughput result we measured.

For this provider, confirm audio handling, resolution, watermark, privacy and commercial permission before committing a paid project. We did not verify a generated download or checkout. Keep the source and compare every transition; an attractive example on the vendor’s page does not predict the outcome for your footage.

First-party sources: VideoExpander outpainting interface

VideoExpander public outpainting interface showing aspect-ratio, padding and quality controls, captured October13,2026.

Public VideoExpander controls captured October13,2026. Interface and vendor claims are not measured output-quality evidence.

Cropping alternative: Premiere Auto Reframe

Adobe Premiere Auto Reframe analyzes footage and adjusts the crop to keep a subject in view across aspect ratios. It is an alternative when important content can fit inside a smaller view; it is not one of the three outpainting services above and does not generate missing scenery.

For an interview with a single centered speaker, an ordinary crop may be more controllable than invented edges. For a wide product shot with two objects at opposite sides, cropping may remove something essential. Decide from the actual shot, not from a rule that generative processing is always better. Inspect tracking around cuts, occlusion and people crossing.

First-party sources: Adobe Auto Reframe documentation

A five-step comparison using your own clip

1. Write the acceptance brief before generating. Specify source and destination ratios, the subject that must remain visible, the permitted new background, exact text or logos that must remain intact, required audio and the intended final placement. State whether cropping or padded borders would already meet the brief.

2. Choose one short permitted clip. Include a moment when the subject moves near an edge and a moment with background detail. Keep an untouched source copy. Avoid confidential or unconsented media while evaluating a vendor. Use the same source section and target shape across the services you compare.

3. Record the actual operation. Keep the source length, trim, target ratio, model if exposed, positioning, prompt, credit quote and output settings. A website’s plan price is not the same as the cost of the particular operation. If an option is not exposed, record that as unknown rather than filling it with a guess.

4. Inspect at full size and through time. Watch normally, then scrub the beginning, end, edge crossings and camera moves. Check new geometry, contact shadows, reflections, repeated objects, texture flicker and any visible boundary. Also compare the original subject and soundtrack. A sharp thumbnail can hide an unusable transition.

5. Export and put the file in its intended layout. Add the actual caption and branding in an editor, not as invented text inside the scene. Check the destination’s safe area, frame rate, compression and audio sync. Count every rejected attempt and cleanup minute before deciding which route is cheapest for work you can use.

Worked example: turning a product clip into a vertical ad

Suppose a five-second landscape clip shows a bottle on a table, with its label centered and the full tabletop visible. The destination is a vertical ad with a headline above the bottle and a CTA below. This is an illustrative production brief, not a result generated for this article.

First test an ordinary vertical crop. If it removes the product, required hands or a useful demonstration, consider expansion. Describe neutral surrounding tabletop and wall space, preserve the bottle’s shape and label, and leave room for typography. Keep the actual headline and CTA as editable overlays after the image is accepted.

Review the bottle neck, cap, label, reflections and contact shadow on every important frame. A plausible new wall is not sufficient if the product itself changes. If a new hand, extra bottle or moving background object appears, reject the take or revise the brief. Do not approve the file merely because it fills a 9:16 canvas.

Once the image is acceptable, restore or confirm the intended audio, place the designed text, and view the ad at phone size. If maintaining every detail requires extensive repair, a padded or designed background can be a more reliable deliverable. Record that decision as part of the comparison instead of forcing outpainting onto unsuitable footage.

Budget by accepted export, not the free label

A useful budget counts generated attempts, manual cleanup, editing and review. For example, four attempts to obtain one usable clip cost the sum of all four displayed charges, plus any cleanup. This is arithmetic for planning, not an observed acceptance rate. Record what happened in your own comparison before generalizing.

Guest allowances, account credits, paid subscriptions and API billing are separate. Do not copy one provider’s credits-per-second number into another product’s estimate. A monthly plan also buys access over time; it is not necessarily the incremental cost of one export. Read the selected tool’s quote, then check the watermark and commercial-use requirements of the file you need.

For repeated production, keep a small log with input type, aspect-ratio change, attempts, rejected reasons and accepted files. Compare similar shots. A locked camera against a simple wall and a fast crowd shot are different populations; mixing them into one success percentage hides the conditions that determine whether expansion is useful.

Common failures and the next useful action

Flickering edges: reduce the amount of new area, test a shorter uninterrupted shot and inspect the camera move. Change one variable at a time. Repeating an identical request does not establish that the next output will fix the problem.

Wrong objects or altered people: simplify the surrounding-scene brief and check whether the chosen source is too crowded or obscured. If preservation is essential, use cropping or a designed border. Do not present generated background objects as things that were actually recorded.

Audio missing or out of sync: compare the exported duration and cadence with the original, then reattach the appropriate original track in an editor. A provider’s visual outpainting claim is not proof that audio survived the operation.

Upload or ratio rejection: use the tool’s displayed supported file and trim settings. In Magic Hour’s Dashboard, the guide calls for a changed destination ratio and complete Custom values within its supported range. Keep the project ID and the exact error if support needs to inspect it.

A job is still processing: check its final status before creating duplicates. A browser timeout and a completed low-quality file are different cases. Preserve the project and source until the issue is understood; follow the provider’s documented support and credit process.

Which tool should you choose?

Start with Magic Hour when you need a short browser trial and a straightforward media workflow. Choose Luma Reframe when canvas placement and prompt-directed surroundings are central. Evaluate VideoExpander when a dedicated outpainting interface is the main requirement. Choose Auto Reframe or designed padding when the missing scenery does not need to be generated.

The strongest choice is the one that delivers your accepted file with the least repair while meeting your permissions and export requirements. This comparison supplies a shortlist and an explicit way to evaluate it. It does not establish that any provider is universally first on realism, speed, price or preservation.

Continue with the right video workflow

For general edits, compare AI video editors. To add duration, use the AI video extender comparison. For more pixels, compare video upscalers. These guides own different tasks from spatial expansion.

Frequently asked questions

No. Frame expansion changes the visible canvas. A video extender adds duration; an upscaler increases pixel dimensions.

Magic Hour’s public page currently lists three guest expansions per day up to five seconds each; free videos may be watermarked. Other providers have their own input and export limits. Check the actual finished-file requirement.

No. Generated surroundings are inferred content. They can be useful creatively but should not be treated as recovered documentary evidence.

Check the selected workflow and actual export. Luma’s newer Ray3.2 FAQ says Reframe preserves original audio; its older guide describes visual-only output. Retain your original track and verify timing after processing.

Runbo Li
Runbo Li
CEO of Magic Hour
Runbo Li is the Co-founder and CEO of Magic Hour, where he builds AI video and image tools for content creation. He is a Y Combinator W24 founder and former Data Scientist at Meta, where he worked on 0-1 consumer social products in New Product Experimentation. He writes about AI video generation, AI image creation, creative workflows, and creator tools.
View author →

Continue Reading

AI video extenders guide with an illustrative timeline showing original footage followed by newly generated frames.

6 best AI video extenders: inputs, limits and costs

6 best AI video extenders: inputs, limits and costs
AI Video Upscalers

7 best AI video upscalers, models & enhancers (2026)

7 best AI video upscalers, models & enhancers (2026)
AI Video Editing Tools

8 best AI video editing tools (2026): choose by workflow

8 best AI video editing tools (2026): choose by workflow
AI Video Generators for YouTube Shorts

6 best AI video tools for YouTube Shorts (2026)

6 best AI video tools for YouTube Shorts (2026)
AI Tools for Instagram Reels

6 best AI tools for Instagram Reels by workflow

6 best AI tools for Instagram Reels by workflow
Cut-paper media workshop comparing video containers, aspect ratios, codecs, and compression for YouTube

Best video format for YouTube: MP4, H.264 & bitrate

Best video format for YouTube: MP4, H.264 & bitrate