One 16:9 clip cropped to 9:16 and 1:1: $0.04, not 3 generations

Crop a 16:9 clip into 9:16 and 1:1 with two video-filter jobs and the total is $1.915, against $5.625 for generating all three shapes. A 15-second Wan 3.0 clip at 720p costs $1.875; each filter encode costs $0.02. The saving is $3.71. The catch is that a crop removes part of the frame.

Two routes

Wan 3.0 lists at $0.10 per output second at 720p on Alibaba Cloud (read 2026-10-07), and Sume bills list times 1.25, which is $0.125 per second. The video filter takes one clip hosted on media.sume.com and returns a new MP4. Its crop op takes fractions of the frame.

What a crop gives up

A 9:16 crop of a 16:9 frame keeps a vertical strip that is about 0.316 of the width: 9 divided by 16, divided by 16 divided by 9. A 1:1 crop keeps 0.5625 of the width. Everything outside the strip is gone, so the subject must sit in the middle third of the frame, or the crop cuts it.

A generated 9:16 clip is composed for that frame. If the action moves across the screen, a crop will lose it, and the cheaper route will cost you the shot. Check one crop before committing the batch.

• Crop works when the subject is centered and still.
• Generate natively when the subject moves across the frame.
• Run the free filter check before the paid encode to catch bad fractions.

Scaling to a batch

Take 20 hero clips. Generating three shapes for each costs $112.50. Generating one and cropping two costs $38.30. The difference is $74.20, which funds a lot of retakes. The condition is that the 20 clips are all framed for the center.

Framing for the crop

If you know you will crop, say so in the prompt: ask for the subject centered with empty space left and right. The crop rectangle is set as fractions of the frame, with x and y in the range 0 to 1 and width and height from 0.05 to 1, so a centered 9:16 strip uses an x of about 0.342 and a width of about 0.316 at full height. Run the free check on that program first, then the encode.

The same logic at other lengths

The filter fee is flat, so the saving grows with clip length. A 5-second 720p clip costs $0.625, and two crops cost $0.04, so cropping saves $1.21 against two new generations. A 30-second clip costs $3.75; two crops save $7.46.

For very short clips the saving is small and the framing risk is the same, so generating natively is a fair choice below a few seconds.

A rule

If the clip is centered, crop. The two filter jobs add $0.04 to a $1.875 clip, under 3 percent. If you generate natively, one 9:16 clip is $1.875 and a second shape another $1.875. See finishing costs for holiday ads and Wan versus Omni for related numbers.