Keep a whole 16:9 shot in 9:16: blur fill instead of cropping
Cropping loses 68% of a 16:9 frame. Timeline fit blur keeps the full shot centered over a blurred copy. How it works on Sume and when to prefer it.

A 16:9 to 9:16 crop throws away about 68% of the width, so wide shots with two subjects or a product and a hand do not survive it. The alternative in Sume Timeline is fit: blur, which keeps the whole 16:9 frame and fills the rest of the 1080x1920 canvas with a blurred version of it.
Luma lists a Reframe feature for Ray 3.2; Sume does not run that model, but the Timeline fit modes give a reframe that never cuts anything out.
The four fit modes
Each video[] slot in Timeline 1.0 has a fit of cover (the default), contain, stretch or blur. The default output is a 1080x1920 MP4, so a wide clip needs a decision.
| fit | What happens | Cost to the shot |
|---|---|---|
| cover | fills the frame and crops the sides | keeps about 32% of the width |
| contain | fits inside, empty bands remain | all of the shot, small picture |
| stretch | distorts to fill | all of the shot, wrong proportions |
| blur | fits inside over a blurred fill | all of the shot, no empty bands |
The arithmetic
A 16:9 shot fitted by width inside 1080 px is 1080 wide and about 608 px tall, which is 32% of the 1920 px canvas height. The other 68% is the fill. With contain that area is empty; with blur it is a blurred copy of the shot.
So blur gives a small subject on a large canvas. For a close-up, the crop in video filter keeps the subject larger. For a wide product shot, blur keeps it all.
A one-slot render
The render needs a spine; use silence mode if the clip carries its own message, and the clip's own length as the duration. Replace a.mp4 with your clip, imported to media.sume.com.
curl -X POST https://api.sume.com/v1/timeline-1.0/render \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: blur-fill-001" \
-d '{"audio":{"mode":"silence","duration_seconds":8},
"video":[{"source_url":"https://media.sume.com/artifacts/artf_demo/a.mp4",
"start":0,"duration":8,"fit":"blur"}]}'When to choose which
Pick blur for wide establishing shots, landscapes, product lineups and two-person frames. Pick a crop for a single subject you can place with one x value. Pick cover when losing the sides is fine. Do not pick stretch unless the distortion is the point.
Compose is a different tool: timeline compose puts one still and one video in the same frame, with its own fit options (cover, contain, stretch, and blur is not a compose fit).
The render costs $0.10 per ceil(output minute). A clip of 8 seconds is $0.10. There is no provider inference in the step.
Compare the three outputs
Run the same 8-second wide clip three ways and look at the results on a phone. A crop with video filter (x 0.3418, width 0.3164) keeps the center and enlarges it. A cover slot does a similar center crop without a separate filter job. A blur slot keeps everything and shrinks the shot to about 608 px of height in the 1920 px canvas.
The crop path costs the extra $0.02 filter job on top of the render. The cover and blur paths cost only the render, $0.10 per ceil(output minute).
Mixed sources in one cut
fit is per slot, so one render can mix a blur-filled wide shot with a cover-fitted close-up. The default output is 1080x1920, so a vertical AI clip passes through at full size, while 16:9 clips need a choice. Keep one rule per scene to avoid the look of a different fit at every cut.
Limits to remember
Timeline output edges are even integers from 256 to 2160, so 1080x1920 is inside the range. The render is a re-encode, so run it once at the end rather than stacking renders. A clip with its own burned-in captions near the edges is a reason to prefer blur over cover, because cover cuts the sides.
Sources
Related posts
More in Media tools
- Kling Motion Control character_orientation: video or image?
Pick 'video' (the default) for complex motion up to 30 s; 'image' follows camera movement but fal documents a 10 s limit. Sume bills $0.1575 per second.
- Kling Motion Control prompt: appearance only, motion is the video
On Sume's Kling 3.0 Motion Control the prompt guides appearance only. Motion comes from motion_video_url, and the body rejects reference_image_urls and model.
- Lip-sync a saved avatar with H3 Max: avatar_handle, not image_url
Use a ready Sume avatar as the face for MiniMax H3 Max Lip Sync: send avatar_handle plus Sume-hosted audio, 5-14.8 seconds, and see the price at 768p.
- Put a logo or lower third over a talking avatar video with compose
Timeline compose overlay places a Sume-hosted still over an avatar clip, at the top, center or bottom. $0.02 flat per job. The layout keys and a request.
Written by Sume