Doubao Seedream 5.0 Lite API
ByteDance Seedream 5.0 Lite generates images from text, edits from references, and produces related multi-image sets through one public model ID. Roko API currently exposes it through the synchronous image-generation endpoint, billed per output and resolution.
- Public model ID
doubao-seedream-5.0-lite- Output tiers
2K / 4K- Billing
Per output and resolution- Generation endpoint
POST /v1/images/generations- Catalog snapshot
What is Doubao Seedream 5.0 Lite?
Seedream 5.0 Lite is the lightweight tier of ByteDance's Doubao Seedream image family, served through the Volcengine Ark platform. Volcengine highlights stronger instruction following, cross-image consistency, Chinese and English text rendering, and professional-scene quality, positioning it as a cost-effective workhorse for production imagery.
It is also the first Seedream model with native web search: the official documentation says the model decides from the prompt whether to fetch live web information and weave it into the result. The upstream model ID is doubao-seedream-5-0-lite-260128.
Roko API currently exposes 2K and 4K output tiers, up to 10 reference images, and sequential sets of up to 15 images. Treat the Roko catalog rendered on this page as the source of truth for available parameters and pricing; upstream capabilities are not automatically open here.
Core capabilities
Text to image and editing
Generate from a text prompt in Chinese or English, or edit and restyle an existing reference image.
Multi-reference fusion
Blend subjects, outfits, or styles from several reference images into one new picture for try-ons and character extensions.
Sequential image sets
With sequential_image_generation=auto, one request returns a related set of images — up to 15 per set.
Built for volume
2K output starts at $0.03 per image, sized for social assets, A/B concept drafts, and batch production.
Doubao Seedream 5.0 Lite API pricing
Pricing is generated from the current Roko model catalog. The total is outputs multiplied by the selected resolution rate; confirm the estimate before submission.
| Resolution | Price | Unit |
|---|---|---|
| 2k | $0.03 | per output |
| 4k | $0.036 | per output |
Cost examples
For social posts, concept drafts, and everyday batch assets.
Sets are billed per delivered image; failures are not counted as generated.
Parameters and supported values
These fields come from the current public generation_config. Volcengine also documents a 3K tier and up to 14 references upstream; parameters outside the Roko form are subject to request validation.
| Parameter | Values | Notes |
|---|---|---|
prompt | textarea | Describe the image to create or the changes to apply to references. · Required |
image | ≤10 | Reference images may use URL or Base64. State what must stay and what may change. · Optional |
size | text | Selects an aspect ratio, resolution tier, or explicit pixel dimensions as configured for the model. · Optional |
sequential_image_generation | disabled / auto | Set to auto to let the model generate a related set of images; disabled produces a single image. · Default: disabled · Optional |
sequential_image_generation_options.max_images | 1–15 | Maximum images in a generated set; reference count plus output count stays within 15. · Optional |
output_format | png / jpeg | File format of generated images: png or jpeg. · Optional |
response_format | url | Synchronous Seedream delivery currently accepts url only; results are copied to Roko-managed storage. · Default: url · Optional |
watermark | boolean | Whether to add an “AI generated” watermark to outputs. · Default: false · Optional |
stream | false | Synchronous Seedream result storage currently requires false; stream=true is rejected during request validation. · Default: false · Optional |
Call Seedream 5.0 Lite with Roko API
Call the synchronous image-generation endpoint; it returns data[].url after result storage completes. Never expose an API key in client-side code.
Generate an image
curl https://api.rokoapi.com/v1/images/generations \
-H "Authorization: Bearer $API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "doubao-seedream-5.0-lite",
"prompt": "An energetic magazine-cover portrait, sculptural hat, rich color blocking, studio lighting, medium-format look",
"size": "2K",
"response_format": "url"
}'A successful response returns images in data[].url and includes a task_id. If result storage temporarily fails and a recovery task_id is returned, poll it as described in the image-task documentation instead of submitting the generation again.
Common use cases
Social and e-commerce assets
Produce post images, product scenes, and selling-point graphics at volume with per-image billing.
White wireless earbuds floating on a light gray backdrop, professional product photography, soft shadows, 2K.
Marketing posters with text
Volcengine calls out improved Chinese and English text rendering, usable for poster drafts with titles and subtitles.
Autumn launch event poster, warm geometric color blocks, title and subtitle with date and venue.
Sets and series content
Generate a related set around one subject or style — tutorial steps, sticker packs, and storyboards.
Using the reference character, generate a four-image set: wearing sunglasses, riding a motorbike, wearing a hat, holding a lollipop.
Editing and outfit swaps
Swap clothing, backgrounds, or elements from a reference by stating what stays and what changes.
Replace the outfit of the person in image 1 with the outfit from image 2. Keep the face, pose, and background unchanged.
Important limitations
- Roko currently exposes 2K and 4K tiers; Volcengine also lists a 3K tier and custom pixel sizes, subject to request validation here.
- size accepts a tier or pixel dimensions and defaults to 2048x2048 upstream; aspect ratios from 1/16 to 16.
- Sequential sets require reference count plus output count within 15; Roko currently exposes up to 10 references.
- The upstream web-search tools parameter is not part of the current public Roko form; treat availability as subject to request validation.
- Text, hands, and fine details can still be wrong — review final assets. Output rights depend on your inputs and applicable terms.
Choosing between Seedream models
5.0 Lite balances cost, sets, and streaming; 5.0 Flash is cheaper at 1K/2K only; 4.5 leads on fidelity and multi-reference consistency; 5.0 Pro targets premium single images. Pick by cost, resolution, and task.
All models →Doubao Seedream 5.0 Lite FAQ
What model ID should I use for Seedream 5.0 Lite?
Use the public Roko API model ID doubao-seedream-5.0-lite. Volcengine's versioned ID is doubao-seedream-5-0-lite-260128; requests through Roko should use the Roko ID.
How much does the Seedream 5.0 Lite API cost?
The current catalog lists $0.03 per 2K output and $0.036 per 4K output. Prices can change — confirm the console estimate and final settlement.
Which resolutions and sizes are supported?
Roko currently exposes 2K and 4K tiers. Upstream also documents a 3K tier and explicit pixel sizes (total pixels between 2560x1440 and 4096x4096, aspect ratio 1/16 to 16); unexposed tiers follow request validation.
Can one request generate multiple images?
Yes. Set sequential_image_generation to auto to produce a related set of up to 15 images; references plus outputs must also stay within 15.
Does Seedream 5.0 Lite support image editing?
Yes. Passing reference images switches the task into image-to-image/editing — outfit swaps, background changes, local edits, and multi-image fusion with up to 10 references (the current Roko value).
Is the web search capability available?
Web search is Volcengine's headline feature for 5.0 Lite, but the tools parameter is not part of Roko's current public form. Treat it as subject to request validation rather than a published capability.
Why does the API not return an image immediately?
Roko currently exposes this model through synchronous POST /v1/images/generations. It waits for generation and result storage before returning data[].url. If storage fails but a recovery task_id is supplied, query GET /v1/tasks/{task_id} for recovery status.
Seedream 5.0 Lite vs 4.5 vs Flash — which one?
Use 5.0 Lite for volume and sets, 5.0 Flash for the lowest-cost small images (1K/2K from $0.02), 4.5 for print-grade fidelity and reference consistency, and 5.0 Pro for premium single images.