AI Models
The models behind Visxy
What each model is good at, where it differs from the others, and what it costs to run.
AI Models
What each model is good at, where it differs from the others, and what it costs to run.
The short draft lane next to Seedance 2.5. Try the shot here; move a keeper to 2.5 only when the clip has to run past a quarter-minute.
Credits come with a Generator plan. The exact cost is shown on the Generate button before anything is spent.
Fast is the model you use to find out whether the shot works. It is not a speed knob on Seedance 2.5. On Visxy it is the first pass through a video-library prompt: 9:16, one idea, short enough to throw away.
If the prompt already needs a beginning, a middle, and an end that will not fit a short clip, skip this page and open 2.5.
Trying motion. The Visxy video library is mostly a single vertical shot with a sentence of action. Fast will take that sentence. Use it to see whether the walk, the turn, or the camera travel reads before you spend 2.5 on a long take of a bad crop.
It is also the right model when you already know the clip should stay short. A loop, a gesture, a three-beat look into camera does not become better because you opened the long SKU. Fast's ceiling is the short-clip ceiling. Stay here on purpose.
What it is not: Seedance 2.5 with the quality turned down, or a story machine. If you catch yourself writing "then… and then… and then…" you are already off this page.
Against Seedance 2.5: own product, own ceiling. Animating a Fast draft must not quietly become a 2.5 job — and on Visxy it will not, because they are paired separately. The visible difference you will feel is duration. 2.5 exists because some prompts need more time than Fast can give. If your prompt does not, 2.5 is the wrong tab.
The provider ids underneath say seedance-2-fast — Fast is the short Seedance lane, labelled 2.5 Fast in this catalog so it sits next to 2.5 in the picker. It is still not 2.5.
Fast's image-to-video still lets you pick a frame, including one that follows the source. Seedance 2.5's animating shape takes the frame from the picture and offers no picker. If you need to reframe a still before it moves, you are on Fast.
Paste the library prompt. Cut any second event. "She turns and then walks away" is two shots; Fast will average them.
9:16 if it came from the video library. Use a wide frame only when the brief is actually wide.
Keep the clip short even when the slider will go further. Drift is what extra seconds do when there is no new action to fill them.
Try the smaller size, then re-run a keeper larger. Same habit as 2.5; here it is how you avoid a long wait on a prompt you will rewrite.
Generated audio is fine for room tone. Dialogue and cues: silent, then your own track.
No Visxy run is on the board yet. Treat Fast as the try, not as the deliverable, until you have seen one clip land.
No public Fast Edit page. Attach a picture to this model and describe the motion only.
If you need to reframe a still before it moves, you are on Fast. If the still is already the crop, either model will take it — use 2.5 only when the move must run long.
A second still is the last frame. Crop before you attach.
No. Fast stops at a short-clip ceiling. A prompt that needs a long single take belongs on Seedance 2.5. Re-run the same text there; do not pad the Fast prompt with empty seconds.
No. Separate catalog row. Fast is the short Seedance. 2.5 is the long take.
Start here when you do not yet know the motion is right, or when the clip should stay short. Start on 2.5 only when the prompt is a sequence Fast cannot finish.
Attach it to Seedance 2.5 Fast and write the motion. There is no separate Edit article. Fast can reframe that still; Seedance 2.5's image-to-video cannot.