Seedream 4.0: lock this still before you buy motion (3cr)
Seedream 4.0 is the stills row. Animate after approval. Do not generate the pack inside the video model.
Seedream 4.0 is the stills row. Animate after approval. Do not generate the pack inside the video model.
Seedream 4.0 is ByteDance text-to-image: no still required, 3 credits (a discounted catalog row), high-resolution generation up to 4K, multi-image output, built-in prompt enhancement. There is no duration because nothing moves. The text-to-image tool is the door. The video model is not a faster way to get twelve product frames.
Stills are the cheap lock
A clip will invent motion you cannot un-invent. A still you can reject in a glance. Seedream 4.0 exists so composition, label, and palette get signed off while the generate is still a picture. Quality settings listed are 1K, 2K, and 4K. Aspect options run 21:9 through 9:16, plus auto. Styles listed are Realistic, Artistic, and Anime. Use those knobs here, not inside a video prompt.
Built-in prompt enhancement will thicken a thin sentence. That is useful for a first pass. It is not a substitute for naming the product, the surface, and the crop. If the enhancement adds a second bottle, you catch it on the still. You do not catch it after you have paid for motion.
Multi-image is the pack, not a clip
The catalog calls out multi-image output. That is how you explore a pack: several stills from one brief, pick the winner, kill the rest. It is not "almost video." Twelve stills that agree are a storyboard. Twelve video takes that almost agree are a credit sink.
Do not ask a video row to search for the hero angle. Search here. When the still is locked, hand it to image-to-video or a reference stack. The reusable characters and products flow is how that still becomes a kit instead of a one-off download.
3 credits before the video row
Three credits is the catalog price for this generate. ByteDance marks the row discounted; the number you plan against is still this stills price, not the video family's. Retrying a 3-credit frame is how you protect a 25-credit clip. Generating the pack "in motion so it feels real" is how you discover the label was wrong on take seven.
No image is required. This is text-to-image. If you already have a plate, you are on an edit row, not this one.
Approve, then buy motion
The test is boring on purpose. Would you print this frame? If not, do not animate it. Seedream 4.0 is the ByteDance stills lock: up to 4K, multi-image, prompt enhancement, 3 credits. Motion is a different catalog row, on purpose. The prompting guide for images covers how to specify the frame. This page is only the order of operations.
FAQ
Is Seedream 4.0 a video model?
No. Content type is image. Category is text-to-image. There are no supported durations. Animate after the still is approved.
What does "up to 4K" mean if max output resolution is blank?
The catalog description is high-resolution generation up to 4K, and quality settings list 1K, 2K, and 4K. Use the quality control. Do not infer a video 4K pass from that.
Why not generate the pack inside the video model?
Because the pack is stills work. Multi-image output is how Seedream 4.0 explores variants. A video generate spends more and hides the bad label inside motion.
Do I need a reference image?
No. This row does not require an image. If the job is "change this locked still," pick an edit model instead of re-rolling Seedream.