A strong video does not always arrive in the shape you need. A landscape product clip may need to become a vertical Reel. A 16:9 campaign video may need a 4:5 feed version. A square asset may need to fill a widescreen presentation. The usual fixes—cropping, stretching, or adding empty bars—can damage the composition.
OpenCake now includes LTX 2.3 Reframe in AI Models for this exact job. It changes the canvas to a new aspect ratio and generates scene-matched content around the source video, helping preserve the original subject and composition instead of destructively cropping them.
Ready to reframe a clip? Open AI Models in OpenCake
What does changing a video’s aspect ratio mean?
Aspect ratio describes the relationship between a video’s width and height. A 16:9 video is wide, a 9:16 video is tall, and a 1:1 video is square. Changing that ratio changes the shape of the frame, not just the number of pixels in the export.
Traditional resizing often zooms in until the new canvas is filled, cutting off whatever falls outside it. LTX 2.3 Reframe takes a different approach: it keeps the source video as the visual foundation and fills newly exposed areas with generated content that follows the scene, motion, lighting, and camera context.
Cropping vs stretching vs generative reframing
| Method | What it does | Typical result |
|---|---|---|
| Crop | Cuts pixels from the top, bottom, or sides to fill a new frame. | Fast, but subjects, products, text, or important action can disappear. |
| Stretch | Forces the original image into a different frame shape. | Keeps everything visible but distorts people, products, and geometry. |
| Padding | Adds black bars, color, or a blurred copy behind the original. | Preserves the clip but often looks like a workaround. |
| LTX 2.3 Reframe | Expands the canvas and generates plausible content for the new areas. | Preserves more of the original composition while creating a frame that feels intentional. |
How to change a video’s aspect ratio in OpenCake
- Open AI Models from the OpenCake dashboard.
- Choose Utilities, then select LTX 2.3 Reframe next to tools such as Remove Background and Image Upscaler.
- Add your source video by uploading a file or choosing an existing clip from your Library.
- Choose the target aspect ratio: 1:1, 4:5, 5:4, 9:16, or 16:9.
- Choose 720p for a faster, lower-credit test or 1080p for a higher-detail final version.
- Review the credit quote. Reframing cost is based on the source video’s duration and selected resolution.
- Generate the new version. The finished video is saved to your OpenCake Library for review, download, or reuse.
LTX 2.3 Reframe does not require a text prompt in OpenCake. The source video, target aspect ratio, and resolution are the only creative inputs, so the workflow stays focused on adapting an existing clip rather than redesigning it through instructions.
Which aspect ratio should you choose?
| Aspect ratio | Frame shape | Common uses |
|---|---|---|
| 9:16 | Vertical | TikTok, Instagram Reels, YouTube Shorts, Stories, and vertical ads |
| 16:9 | Widescreen | YouTube, websites, presentations, connected TV, and landscape ads |
| 1:1 | Square | Social feeds, product posts, marketplace media, and flexible ad placements |
| 4:5 | Portrait | Instagram feed posts, Meta feed ads, and mobile-first product creative |
| 5:4 | Landscape | Landscape feed placements, portfolio media, and compact horizontal layouts |
Example: turn a 16:9 video into a 9:16 vertical clip
Suppose you have a wide product video with the product in the center and useful scenery on both sides. A normal 9:16 center crop may keep the product but remove the environment, supporting action, or camera movement that makes the shot work.
In LTX 2.3 Reframe, select the source clip, choose 9:16, and run a 720p test first. The model creates a tall composition around the source scene rather than simply slicing away the left and right sides. If the motion and generated regions look clean, repeat at 1080p for the higher-detail version you plan to publish.
When to use 720p and when to use 1080p
Use 720p while testing several source clips or deciding which aspect ratio works best. It processes at a lower cost per input second and is usually enough to judge composition, continuity, and edge generation. Choose 1080p when the result will be delivered, published, shown full-screen, or reused in a high-quality production workflow.
A practical workflow is draft at 720p, inspect the full clip, then generate only the approved direction at 1080p. This avoids spending final-quality credits before you know whether the new composition works.
How to get better reframing results
- Start with the cleanest source available. Compression artifacts, blur, and unstable footage give the model less reliable scene information.
- Choose clips with a clear subject. Products and people that remain visible and reasonably centered are easier to preserve across very different frame shapes.
- Test the biggest ratio change at 720p first. Moving from 16:9 to 9:16 exposes much more generated area than moving from 16:9 to 5:4.
- Watch the whole output, not only the first frame. Check whether backgrounds, reflections, shadows, and moving objects stay coherent over time.
- Inspect faces, hands, logos, packaging, signs, and text near the expanded edges. Generated regions can look plausible while still introducing incorrect details.
- Keep a safe version of the source. Reframing creates a new asset in the Library, so your original remains available for other formats.
What to review before publishing
Generative reframing avoids destructive cropping, but it still creates new pixels. Treat the result as an edited video that needs review. Look for objects that appear or disappear, duplicated people, shifting architecture, warped reflections, inconsistent light, and any invented brand detail in the expanded parts of the frame.
Also preview the video in the destination layout. Platform interface elements can cover the top, bottom, or sides of a clip. Keep important products, faces, captions, prices, and calls to action inside the safe area even when the generated background fills the entire canvas.
Best use cases for LTX 2.3 Reframe
- Convert a landscape product commercial into a vertical social ad.
- Adapt one campaign master to 9:16, 4:5, 1:1, and 16:9 placements.
- Recover more usable composition when a traditional crop removes a person or product.
- Turn horizontal founder, tutorial, or UGC footage into a mobile-first version.
- Create widescreen versions of vertical clips for websites, presentations, or YouTube.
- Prepare multiple aspect-ratio variants before adding final captions or platform-specific graphics.
Frequently asked questions
Can I resize a video without cutting anything off?
LTX 2.3 Reframe is designed for that goal. Instead of filling a new frame by zooming and cropping, it generatively expands the scene around the source. You should still review the result because the added areas are model-generated rather than recovered from the original recording.
Can I convert a horizontal video to vertical?
Yes. Choose 9:16 for a full vertical frame or 4:5 for a portrait feed frame. Both preserve more of the source composition than a simple center crop by generating content around it.
How long can the source video be?
The OpenCake LTX 2.3 Reframe route supports source videos up to 60 seconds. Longer clips should be divided into intentional shots before reframing so each result can be reviewed for continuity.
Does changing the aspect ratio lower quality?
Any generated expansion can introduce differences around the original frame. Choose a clean source and 1080p output for final delivery, then review motion and fine details. Use 720p for composition tests rather than assuming the first result is publication-ready.
The fastest workflow
Open AI Models, choose LTX 2.3 Reframe under Utilities, add a video, select the new aspect ratio, test at 720p, and generate the approved version at 1080p. The result returns to your Library, ready for captions, cleanup, download, or the next stage of your OpenCake workflow.