SanShengAI
登录开始生成

REFERENCE IMAGE PREP MODEL

Seedream 5.0

Clean still references, product frames, and video reference prep before you animate in a separate model.

Use Seedream as an image generation and editing workspace for source stills: create product frames, character references, ad visuals and controlled image sets before sending the strongest assets to a video model.

Prepare references with SeedreamAnimate with Seedance
Open image workspaceView pricingPrompt examples
Seedream product reference still prepared for a later video workflow
Image model
Up to 10 refs2K-4K

Seedream reference example

Still reference prepared for later animation

Still reference prep

Create clean source images before handing assets to a video route.

Text or image input

Start from a prompt or edit with uploaded reference images.

Up to 10 references

Guide edits with multiple source images when details need tighter control.

2K or 4K images

Prepare higher-resolution stills for product, character and layout review.

Up to 4 outputs

Generate up to 4 stills per request on the current Seedream route.

Per-image pricing

Image presets use the shared pricing helper, not video seconds.

Seedream image pricing at a glance

Per-image preset totals - see the exact live price in the image workspace before you generate.

View full pricing

2K image

US$0.10

Single generated still

4K image

US$0.20

Most popular

High-resolution still

Reference batch

US$0.40

4 generated images · US$0.10/image

All prices are SanShengAI display prices in USD credits for preset image scenarios.

Seedream 5.0 examples

Explore real community outputs and see how Seedream 5.0 performs inside MaxVideoAI workflows.

Still image examples

Prompt patterns tailored to Seedream 5.0.

Typography control

Exact copy, hierarchy and placement stay explicit.

Reference edits

Keep product identity, palette, layout or style.

4K finals

Validate at 2K, then finish at 4K.

Production-aware

Owned references and built-in guardrails.

When should you choose Seedream?

Choose Seedream when the next video pass needs cleaner product, character, style or composition references first.

Open Seedream image workspace

Preparing product frames?

Generate stills with readable labels, clean backgrounds and stable framing before using them as source assets elsewhere.

Open Prompt Lab

Need motion next?

Seedream prepares the stills. Pick a video model after the reference set is ready.

Browse video models

How to write a strong Seedream prompt

Write the still image as if it were the first frame of a shot: subject, composition, lighting, and constraints first.

Tip: if the image is going into video later, mention the future motion goal without asking Seedream to animate it.

How Seedream prepares references

Prompted still

Describe the exact product, character, scene, surface, lens and lighting for a clean source frame.

Reference edit

Upload source images when the output must preserve identity, product geometry or brand details.

Reference set

Create multiple stills for front, side, detail and lifestyle views before the video step.

Size choice

Use 2K for faster prep, or 4K when the still needs extra detail.

Video handoff

Send only the strongest stills into the chosen motion model once the visual direction is clear.

Global principles

  • Use one clean visual goal per image.
  • For image sets, name the sequence, count, and shared visual rules.
  • For edits, say what to preserve before what to change.
  • Keep the background simple when the still will become a video reference.

Engine quirks / what to watch for

  • Controlled stills give image-to-video models clearer starting material.
  • Product and character references are easier to animate when silhouette, lighting, and props are already stable.
  • Drafting the still first helps separate image quality decisions from motion decisions.

Seedream 5.0 - image generation and editing image prompt examples

Seedream prompts for product stills, character reference sheets, image edits and storyboard batches.

Open image workspace

Product reference still

2K still

2K product still of an amber perfume bottle on white acrylic, soft studio light, clean shadow, readable label, minimal background ready for reference use.

Character reference sheet

2-10 refs

Character reference sheet with the same face, outfit and palette across four clean views, neutral background, consistent jacket details and accessories.

Image edit cleanup

Image edit

Using this product image, preserve shape, logo and proportions. Clean dust and unwanted reflections, simplify the environment and keep natural lighting.

Storyboard batch

4 images

Create four coherent launch storyboard images: product hero, material detail, lifestyle context and final frame. Same palette, same lighting, distinct compositions.

Tips and limits

Treat Seedream as the still-image craft step. It can produce better source material, but it does not replace downstream prompt and policy checks.

What works best

  • Generate clean product, character, and environment frames before animation.
  • Generate related image sets for storyboards, product variations, and brand systems.
  • Use text, single-image, or multi-image inputs for creation, editing, and image fusion.
  • Use edit mode to reduce visual noise in existing source images.
  • Keep one clear subject per frame when the output will become image-to-video input.
  • Use the same language for subject, lighting, and style when moving from Seedream into video.

Hard limits to keep in mind

  • Still images only.
  • The current public route supports up to 10 reference images.
  • The current public route supports up to 4 generated outputs per request.
  • Downstream video generation still evaluates each source image and prompt.
  • A cleaner reference image may improve consistency, but final animation handling and output quality still depend on the video run.
  • Do not use Seedream to prepare content that violates SanShengAI safety rules.

Specs

The limits that shape your renders.

View full specs

Price / image

4K US$0.20/image2K US$0.10/image

Text-to-Image

Supported

Image-to-Image

Supported

Resolution options

2K / 4K

Aspect ratios

16:9 / 9:16 / 1:1 / 4:3 / 3:4

Output format

jpeg / png

Seedream image workflows

Use Seedream 5.0 for text, single-image, and multi-image workflows inside the SanShengAI Image Lab.

Details
  • Text-to-image: create a still image from a prompt with 2K or 4K output controls.
  • Image-to-image: clean or edit an existing source image while preserving important visual details.
  • Multi-image input: blend subject, style, or scene elements from up to 10 reference images.
  • Small image sets: create related visuals for product variants, storyboards, or brand systems.

Limits and billing behavior

Details
  • The current public route allows up to 4 generated images per request.
  • SanShengAI prices Seedream by successfully generated image; failed or missing outputs should not be charged as completed images.
  • Output controls expose resolution, aspect ratio, file format, and watermark options where supported.
  • The $0.24 batch preset is an example of 4 generated images at $0.06 per generated image.
  • Output format: choose jpeg or png in the active Image Lab route.

Safety and likeness

Built-in safeguards and best practices for responsible creation with Seedream 5.0 - image generation and editing.

  • Use original characters and owned references.
  • Avoid real people, celebrities and protected characters.
  • Do not use someone's likeness without consent.
  • Avoid copyrighted franchises, logos and protected IP.

FAQ

What is Seedream 5.0?

Seedream 5.0 is an image generation and editing model available in SanShengAI for still-image creation and reference preparation.

How does Seedream connect to Seedance?

Seedream is a recommended reference-image companion before Seedance 2.0. Create or clean a still image first, then use the selected frame as a stronger starting point for a later video pass.

Can Seedream generate small image sets?

Yes. The current public route supports up to 4 generated images per request, which is useful for product variants, storyboard frames, and reference exploration.

Does Seedream promise video results?

No. Seedream can help create cleaner source images, but downstream video generation still evaluates each prompt and source image under its own workflow and safety rules.