Models · two steps to a video
Models
A Breyvi video is made in two steps. First a picture model draws every scene, then a clip model brings each picture to life. Here is every model for each step: what it makes, what it costs, and how it placed when people voted blind.
If you want
Every picture model
The same task for eachThis character, and one sentence:the same beaver in his suit, on a rainy neon street at night, holding a red umbrella
Look for who keeps the beaver.
01Seedream 5 ProFor paid plansByteDance5 cr
02Muse ImageMeta1 cr
03Nano Banana ProFor paid plansGoogle9 cr
04Nano BananaGoogle3 cr
05FLUX.1 Kontext MaxFor paid plansBlack Forest Labs5 cr
06FLUX.1 Kontext ProBlack Forest Labs3 cr
07Ideogram CharacterFor paid plansIdeogram6 cr
08Seedream 4.0ByteDance2 cr
Side by side
Press the plus on up to three cards. The best figure in each row is the bright one.
Places are from blind pairwise votes, as Artificial Analysis published them on November 11, 2026: AA-Video-I2V v2.1 for clips made from a picture and AA-Image-Editing v2.1 for pictures. “Not rated” is a model the arena does not list, not a low score. Prices are Breyvi’s own, in credits, from the registry the editor charges from; what a credit costs is on the pricing page. The pictures were each made once, on the model named, from the character and sentence above.
