AI Models
What each model is good at, where it differs from the others, and what it costs to run.
ByteDance's long-take video model. Use it when a Visxy video prompt has a sequence to hold — not when you only need a few seconds of motion.
Credits come with a Generator plan. The exact cost is shown on the Generate button before anything is spent.
Seedance 2.5 is the model you open when the clip has to hold a beginning, a middle, and an end in one take — not when you only need a few seconds of a walk. ByteDance built it for that longer pass. On Visxy it is the video-library prompt that already describes what happens next, not a loop of the same gesture.
If you are still finding out whether the shot works, start on Seedance 2.5 Fast. This page is the sequence Fast cannot finish.
Order. A Visxy video prompt that says what happens, in order, in one or two sentences is this model's work. ByteDance's own pitch is a single take that can carry setup, turn, and close rather than stretching one moment. That is the only reason to be here instead of Fast: Fast cannot honour a brief that needs that much time.
It is also the model that rewards structure over vibes. "Cinematic, epic, slow motion" will give you atmosphere and no plot. "She crosses the square, the camera tracks, she stops at the stall" is a Seedance 2.5 prompt. The library's 9:16 portraits-in-motion belong here when the prompt actually has beats.
What it is not: a five-second test button, Fast with the quality turned up, or a 4K master other hosts may advertise on a SKU we do not list. Many things moving at once still fall apart. Physical interactions between several people are the gap ByteDance itself flags.
Against Seedance 2.5 Fast: different model, different ceiling. Fast is not a speed setting on 2.5. Fast stops where a short clip stops. If you liked a Fast take of the motion and the only problem is that the story was cut off, re-run the same prompt here. Rewriting it "to be more 2.5" usually loses the beat you already approved.
Animating a Fast draft must not quietly become a 2.5 job — on Visxy it will not, because they are paired separately.
Hosts who sell Seedance 2.0 as the 4K / short-form SKU are talking about a model we do not list here. Do not pick 2.5 because you wanted that file. Pick it because the prompt has duration in it.
Image-to-video on 2.5 takes the frame from the picture and offers no shape picker. Fast's animating shape can still reframe. Crop before you attach here.
Write the events in order. One or two sentences. If a library prompt stacks three camera ideas and two costume changes, cut it to the take you actually want.
Stay on 9:16 when the prompt came from the video library. A wide frame is for a landscape brief, not for "making it cinematic."
Use the length on purpose. A long clip is not a slightly larger version of a short one. Extra seconds with no new action are drift. Draft the motion on Fast; open 2.5 when the keeper actually needs the long take.
Try the shot at the smaller size first, then re-run what works at the larger one. Do not start at the top size on a prompt you have never seen move.
Generated audio suits ambience. If the sound carries meaning — a line, a cue, a joke — generate silent and add your own. We have not published a Visxy timing for Seedance 2.5. Plan for a wait measured in minutes on a long clip, not a still-image refresh.
There is no public Seedance 2.5 Edit page. Image-to-video is this model with a picture attached. Describe only the motion. Re-describing the scene pulls the clip away from the picture you attached.
One still is the opening frame. A second, if you attach it, is the closing frame. Crop the reference to 9:16 before you attach it if that is the delivery.
No. Fast is its own model with a short ceiling. 2.5 is the long-take SKU. If the prompt does not need that extra time, Fast is the page.
Yes, when they already describe a sequence. A 9:16 library prompt that is really one gesture should stay on Fast. A prompt that names what happens next belongs here.
No. Fast stops at a short-clip ceiling. Re-run the same text on 2.5.
No. Attach a reference to Seedance 2.5 and describe the motion. A second still, if you have one, is the last frame. There is no separate public page for that shape.
