ByteDance · BytePlusIMAGECore tiersNative audio

Seedream on VisionX.

Five image variants on flat VX pricing, led by Seedream 5.0 Pro.

5models on rosterfrom 0.6 VXper imageup to 14reference imagesCast-readyidentity locked
01

What Seedream is on the roster

THE MODEL

Seedream is the core image lineup: five variants from the original 3.0 to the flagship 5.0 Pro, all priced at a flat rate per image on the core tiers regardless of which variant renders.

It is also the reference workhorse. Seedream variants accept up to 9 reference images on a single call, which makes them the default route for character sheets, product stills, and any frame that has to match a Cast.

Flagship quality at the lowest cost on the roster

Seedream 5.0 Pro is the cheapest of the top-tier image models VisionX runs, priced the same for one image or a thousand.

Reference-first

Up to 9 reference images per call. Bind a Cast member or a plate and the output stays on-model.

A variant for every stage

Probe on 4.0, produce on 5.0 Lite, master on 5.0 Pro. 4.5 and 3.0 stay available for looks that were built on them.

Frames that come apart into layers

Seedream 5.0 Pro can split a finished still into a base plus up to 16 transparent PNG layers, each named, ordered and boxed, so a packshot becomes parts you can edit, re-reference or rebuild.

02

What Seedream actually does

CAPABILITIES

Why the core image lineup is the default route rather than the cheap one.

Flat pricing regardless of variant

The core tiers charge the same per image whether the frame renders on 5.0 Pro or on an earlier variant, so choosing quality is not a budget conversation.

Flat rate

Reference-first by design

Seedream variants accept a deep stack of reference images on a single call, which is what makes them the default route for character sheets and any frame that has to match a Cast.

Cast consistency

A variant for every stage

Probe on 4.0, produce on 5.0 Lite, master on 5.0 Pro. 4.5 and 3.0 stay pickable for looks that were graded on them.

Draft to master

Stills that become shots

An approved Seedream frame drops straight into a video generation as the opening frame, so key art and finished film share one face without a second brief.

Image-to-video handoff

Split a still into layers

Seedream 5.0 Pro decomposes a finished frame into a base plus up to 16 transparent PNG layers, each named, z-ordered and boxed, so a packshot becomes parts instead of one flat image. It runs as a finishing pass from the Editor, the Library card and the node canvas, on any still the workspace owns.

Layer decomposition

Edits that stay where you put them

Box or point at the region and describe the change; 5.0 Pro reads the coordinates and the Editor composites the result back through your mask, so everything outside stays byte-identical. A decomposed layer goes back through the same loop on its own, transparency intact.

Coordinate-grounded editing
03

Every variant, exactly as the studio runs it

SPEC SHEET

These rows are read live from the engine registry and the pricing engine. What you see here is what the composer quotes.

ModelModesReferencesOutputVX
Seedream 5.0 ProSeedance 2.0 tierImageup to 9 refsstills1.3 VX · per image, no references
Seedream 5.0 LiteSeedance 2.0 fast tierImageup to 9 refsstills0.7 VX · per image, no references
Seedream 4.0Seedance 1.5 pro tierImageup to 9 refsstills0.6 VX · per image, no references
Seedream 4.5Seedream 4.5 tierImageup to 14 refsstills0.8 VX · per image, no references
Seedream 3.0Seedream 3.0 tierImageup to 0 refsstills0.6 VX · per image, no references

Self-serve rates run as low as $0.085 per VX at scale. The composer quotes the exact VX for your shot, references included, before you run it.

04

What teams route to Seedream

USE CASES

Where Seedream carries the campaign.

Character sheets

Build a Cast member out from a few approved frames, then hold that identity across every downstream engine.

Product stills

Packshots and hero product frames that keep geometry and finish intact under a new light.

Key art at volume

Flat per-image pricing means exploring twenty directions costs what exploring one costs, per frame.

Storyboard frames

Boards that already look like the film, because they were rendered by the family that will shoot it.

FAQ

Seedream, answered

Which Seedream variant should I use?
Probe on 4.0, produce on 5.0 Lite, master on 5.0 Pro. Because the core tiers price images flat, the variant choice is about look and turnaround rather than cost. 4.5 and 3.0 remain available for looks that were graded on them.
Why is Seedream the default for reference work?
It accepts the deepest reference stack of the core image routes on a single call, so a Cast member, a plate and a product can all condition one frame. That is what keeps a campaign on-model across hundreds of stills.
Can a Seedream still become a video shot?
Yes. An approved frame drops into a video generation as the opening frame, so the campaign key art and the finished film share the same face without re-briefing a second tool.
Can Seedream split an image into layers?
Yes. Seedream 5.0 Pro decomposes any still the workspace owns, whether it rendered on Seedream, on another engine, or arrived as an upload, into a base image plus up to 16 transparent PNG layers with names, z-order and bounding boxes. Open the Layers module in the Editor, pick the output size, optionally name or box the elements you want separated, and download the stack or send a layer back into an edit, a reference or the canvas. It is a finishing pass with a flat per-still charge, quoted before you run.

Try Seedream on your Cast.

20 VX free on signup, no card required.