MiniMax H3 with 7 reference images and 6 s at 768p: $0.65 on Sume

Seven reference images and a 6-second clip at 768p on minimax-h3 cost $0.65 on Sume: 6 x $0.075 = $0.45 for the video, plus 2 extra images at $0.08 list x 1.25 = $0.10 each, which is $0.20. The first five reference images are free; each additional image is $0.08 at provider list, which Sume bills at $0.10.

The cost by image count

Using a 6-second 768p clip as the base, the video is a fixed $0.45 and only the image surcharge moves.

The limits around it

The model takes up to 9 reference images, up to 3 reference videos and up to 3 reference audio clips (2 to 15 seconds each, 15 seconds combined). All references together cannot exceed 12, and audio cannot be the only reference. Duration is 5 to 15 seconds, and the 768p tier is native, not 720p.

MiniMax H3 Max works differently: the catalog says reference-to-video there is billed as output seconds at list times 1.25, and the provider's reference-token overage is not reserved. For the same 6 seconds at 768p, H3 Max is 6 x $0.10 = $0.60 regardless of image count.

Which row is cheaper

With 9 images on H3, the video is $0.45 and the surcharge is 4 x $0.10 = $0.40, so $0.85. H3 Max for the same clip is $0.60. From seven images up, H3 Max wins on price for a 6-second 768p clip; up to six images, H3 is cheaper ($0.45 to $0.55 against $0.60). Sume lists the reservation as an estimate, so check usage.cost on the finished job.

Reservation versus final

Only the reference surcharge is tied to the number of images; the video rate is tied to output seconds. If you remove an image you do not need, you save $0.10 once you are above five. Review the references before you submit: a sixth image that repeats the fifth costs 10 cents and adds nothing.

The finished job reports usage.cost. Compare it with your arithmetic on the first run, then trust the table for batches.