Skip to main content

Uncensored — This model declares no content filter. SpicyAPI forwards the request without adding a platform content scan; legal and consent duties remain with the customer.Browse every model at this filtering level
bytedance/seedream-5.0-lite/text-to-image

Seedream 5.0 Lite text-to-image is ByteDance's preset-sized tier: a prompt of 3 to 3,000 characters becomes one 2K image. Pick one of eight width*height presets, from 2048*2048 to a 3136*1344 strip, in JPEG or PNG. Uncensored.

From $0.035/image
Run model

Workbench

Input

Text to image
500/3000
3 to 3,000 characters. Write it for the frame you chose — the wide preset wants a composition staged across a strip, not a portrait pushed sideways. at least 3 characters · at most 3000 characters
A named 2K preset, eight options from 2048*2048 square to 3136*1344 ultra-wide. It fixes the frame's shape and tier; the delivered pixels can come in slightly under the named numbers (a 2048*2048 request has returned 1920×1920). Anything outside the enum is rejected.
Output format
JPEGPNG
jpeg by default, or png. png is lossless and the safer handoff into flatting, retouching or overpainting; it changes the delivered file, not the image that was generated.
Safety checkerOff

SpicyAPI does not run a safety checker on this model. Your prompt and media go to the model as written, and the model itself decides what comes back.

Checking your account before enabling submission and pricing.

Output

Ready for a new run
siren-rocks-stripExample result
Edit
Split into layers
Animate

Related models and next tasks

Missing the model you need?

Tell us what to add and where you would use it. Requests go to the people who decide what gets onboarded next.

Eight named 2K presets decide the frame before the prompt

Seedream 5.0 Lite's text-to-image endpoint turns a written brief into one image at a preset size you name. size takes one of eight width*height strings, from a 2048*2048 square to a 3136*1344 strip, and every preset renders at the 2K tier. There is no aspect_ratio field and no seed here, so the preset is the whole framing decision and the prompt text is the only record of how a frame was made.

Open the playground
Eight named 2K presets decide the frame before the prompt
Size presets
8 named width*height strings, 2048*2048 to 3136*1344
Output tier
Every preset renders at 2K
Prompt length
3 to 3,000 characters
Output format
jpeg (default) or png
Seed
None — identical requests return different frames
Content policy
No content filter, no maturity flag
Framing
size is the only framing controlThere is no aspect_ratio field to reconcile against it. Pick one of the eight presets before writing the prompt, because the composition is built for that exact shape from the first token.
Budget tier
The cheaper place to settle a compositionIterate here while the layout is still moving, then carry the same brief over to Seedream 5.0 Pro once the direction is fixed and the render has to ship.
Delivered pixels
Named sizes, not guaranteed pixel countsThe preset fixes the shape and the 2K tier, and the delivered pixels can land slightly under the named numbers. Read size as a shape and a tier rather than an exact pixel count, and measure the returned file when a layout depends on a precise figure.

What the Seedream 5.0 Lite text-to-image API gives you

One brief, each clause doing a single job

One brief, each clause doing a single job

Subject, wardrobe, material, palette and lighting each read as one clause in the same pass, up to 3,000 characters. Swap only the material clause and keep the rest worded the same to compare a matte, a gloss and a metallic finish on the same described object, the way a materials board compares swatches side by side.

The frame is fixed before the prompt is read

The frame is fixed before the prompt is read

size names one of eight width*height presets, and the composition is built for that exact shape rather than cropped into it afterward. A 3136*1344 brief is written for a strip; the same words forced into the square preset just leave dead space at the edges.

  • Illustration holds its line quality at 2K

    Cel shading, flat colour and deliberate rim light survive at these sizes as well as photography does. Request png when the frame goes on into flatting or overpainting; it changes the delivered file and leaves the render itself alone.

  • No separate route for uncensored work

    Portrait photography and character illustration are ordinary prompts on this endpoint. There is no maturity flag, no content filter and no second call to look up before sending the request. What you do have to handle is the rights to what you publish and the lawfulness of it, since nothing here screens a brief on your behalf.

  • A cheap first pass before the Pro render

    size is a named preset and the price sits at the low end of this family, so a composition can be drafted several times here before a call goes to Seedream 5.0 Pro for material detail. The wording carries over unchanged; what changes on the Pro side is that the frame comes from aspect_ratio and a resolution tier instead of a named preset.

  • No seed, so the prompt is the record

    Two identical requests return two different frames. Store the exact prompt text, the preset and the output_format alongside anything meant to be extended into a series later, because the request is all that can be replayed and replaying it produces a sibling frame.

Where Seedream 5.0 Lite text-to-image fits

  • Compare finishes without a reference photo

    Product concepting

    Compare finishes without a reference photo

    Write one brief for the object and its setting, then re-run it changing only the material or colour clause; the rest of the description carries over word for word. There is no seed, so the pose will not repeat exactly between runs; the pair reads as two takes on one concept.

  • Pick the size before writing the headline zone

    Editorial and poster layout

    Pick the size before writing the headline zone

    Choose the wide or tall preset the placement needs, then describe the empty space where a headline will sit. The layout arrives built for that shape instead of a square crop with a title pasted over it later.

  • Assemble a scene from named parts in one brief

    Concept and key art

    Assemble a scene from named parts in one brief

    State the character, the palette and the environment as three separate clauses, each doing one job. A structured brief resolves them into a single coherent frame, where a vague one averages into something generic. The 3,000-character field leaves room to give all three real detail.

  • Costume-led sets without a crop pass

    Illustration and character work

    Costume-led sets without a crop pass

    Generate a whole set of anime or illustration frames at one preset, from a costume change to a background swap, so every image in the set shares a frame shape. A gallery or card layout then takes them directly, with no crop pass between generation and placement.

How to call the Seedream 5.0 Lite text-to-image API

  1. Choose the preset before writing

    size is the whole framing control here. Pick one of the eight presets, square for a feed asset or 3136*1344 for a hero strip, before drafting a single word of the prompt.

  2. Write subject, wardrobe, material and light in one pass

    There is no separate style field, so state everything together within 3,000 characters, matched to the frame: a wide preset wants elements spread across the width.

  3. Set output_format up front

    Pick jpeg for a straight delivery or png when the result is going on into flatting, retouching or overpainting; it does not change the price.

  4. Submit and poll for a terminal state

    The request queues asynchronously and returns nothing inline. Check the task status until it reaches a finished state, then pull the image from platform storage.

  5. Keep the prompt with any frame worth reusing

    There is no seed, so nothing about the request lets you regenerate a match. Store the exact wording next to the output if a series is planned.

Seedream 5.0 Lite text-to-image worked examples

  • A costume set at one shared frame

    Pick a single preset for the whole set so no cropping is needed before it goes into a gallery layout.

    Anime illustration, an adult catgirl maid in a black satin dress with white lace trim, cat ears and tail, holding a silver tea tray, warm afternoon light through lace curtains, soft cel shading, pastel cream and rose palette, detailed linework

  • A banner composed for its strip

    Writing the headline zone as empty sky keeps the layout from needing a crop pass later.

    Wide banner illustration, a lone astronaut walking across red Martian dunes at sunset, long shadow trailing behind, dusty orange and violet sky, the upper third left as clean empty sky for a title, cinematic wide composition

  • A finish comparison from one brief

    Only the material clause changes between the two runs; everything else in the prompt stays the same wording.

    Product still life, a ceramic pour-over coffee dripper on a linen cloth, soft studio light from the left, matte stoneware glaze, plain studio backdrop

Seedream 5.0 Lite text-to-image questions

State what the frame is for before writing what is in it, because a feed square and a banner strip need different staging. Then describe subject, wardrobe, material, palette and light in one pass of up to 3,000 characters, matched to the preset already chosen. Store that wording with the output: there is no seed, so the text is the only part of the request that can be reused.

All Seedream models5