Seedream 5.0 Lite is an image generation and editing model in ByteDance's Seedream series, suitable for progressing from text-based ideas and reference images to a series of visual assets. doubao-seedream-5-0-lite-260128 supports high-resolution output, related image sets, streaming delivery, and web search, focusing on continuous creative needs rather than transparent-background and layer-separation tasks.
Submit requests to the public API at api.acedata.cloud using the documented parameters, then use the results in your application.
Input parameters and result formats vary by service. Use the public API for this model and follow its guide for generation, task retrieval and editing operations.
The values and options above describe this platform's supported invocation range for this model, and do not represent the model's native capacity or the actual number of outputs per task.
Lite supports related image sets, generating multiple content-related images around the same theme. Prompts can specify shared characters, visual style, and variations for each image at the same time, then enable automatic image sets. Compared with generating each image independently, this approach is better suited for organizing storyboards, product series presentations, and coordinated content illustrations.
In addition to describing an image directly, you can input one or more reference images and use text to explain what should be preserved and modified. The workflow can be used for background adjustments, style transformations, or scene redesign. Clearly specifying the subject, composition, and scope of changes helps each round of editing progress toward a defined goal.
2K, 3K, and 4K presets make it easy to select output tiers for different assets, while explicit dimensions are used to specify the canvas format. Series tasks can receive per-image success or failure events through streaming, then obtain the completed results; the web search tool can also be enabled per task, providing an option for creation that requires retrieved information.
Provide product images and describe the target scene, color tone, shooting angle, and series variations to generate multiple presentation directions. You can first use a lower setting to select compositions, then generate higher-resolution assets for the chosen options. The deliverables are suitable for product scene proposals, campaign illustrations, or advertising visual drafts.
Break the story into clear visual segments, unify character appearance, environments, and art style in the prompts, and describe actions and camera changes separately. Enable related image groups to obtain a series of images for layout, review, and subsequent revisions. Reference images can help convey character settings, but continuity still needs to be checked frame by frame.
Combine article topics, visual requirements, and output dimensions into generation tasks, return image links or Base64 data, and integrate them into the content production workflow. Use streaming when images need to be displayed one by one; use asynchronous queries or callbacks for background processing, and finally archive images by task identifier to avoid making the frontend wait for complete results.
If the task requires related image groups, streaming delivery, or web search, 5.0 Lite is better suited to these workflows. If the core requirement is transparent background control or separating independently editable layers, choose 5.0 Pro. The two are not simply differentiated by resolution: Lite provides 2K/3K/4K, while Pro uses 1K/1.5K/2K for standard generation.
Seedream 4.5 also supports sequential image generation and streaming output, so existing workflows of this type do not need to switch solely because of image group requirements. Lite's specific value lies in adding a 3K option, web search, and JPEG/PNG file format control. It is recommended to compare actual assets using the same prompts and reference images before deciding whether to migrate.
Clearly specify the purpose of each image, subject retention requirements, and target copy; use image to submit an image URL or Base64 for editing, and avoid uploading assets without revision instructions.
Specify model=doubao-seedream-5-0-lite-260128, prompt, and size=2K to /seedream/images. To generate related image groups, set sequential_image_generation=auto and use max_images to limit the maximum; choose streaming and asynchronous/callback modes separately.
Read the images actually returned rather than inferring the number from the request limit; synchronous, streaming, and asynchronous modes each use their own response formats, while asynchronous tasks are linked through /seedream/tasks.
Create three city guide images in the same style based on the provided attraction information: one introducing architecture, one introducing a walking route, and one introducing local cuisine; use only verified short labels for all text.
Provide accurate information first; if web_search is explicitly used, still verify attraction names and geographic relationships, then check the text and actual number of each image.
Send a POST request to /seedream/images, use the full model value doubao-seedream-5-0-lite-260128, and provide a prompt. For editing tasks, also include image. Do not use abbreviations such as doubao-seedream-5.0-lite in place of the full model name.
image can use an image URL or Base64; submit multiple images as an array. The prompt should clearly specify the role of each reference image, such as the subject's appearance, composition, or style, and explain which parts need to be preserved and modified. This makes it easier to organize the task than simply uploading images without stating the editing goal.
Set sequential_image_generation to auto, and specify the series theme and variations for each image in the prompt; use max_images to control the maximum number. If you only want normal single-image output, use disabled. Final delivery is based on the actual returned image results.
When you need a preset resolution, use 2K, 3K, or 4K, and describe requirements such as landscape or portrait orientation in the prompt. When you need specific width and height, use a pixel-dimension string, but it must meet the model's size range. Do not assume that 4K means any width-height combination can be generated.
When stream is enabled, use NDJSON to receive per-image success or failure events, and wait for the final completion event. For background tasks, use async to obtain task_id, then query /seedream/tasks, or use a callback to receive results; these methods cannot be used together with streaming.
Choose a development approach or application suited to this model and follow its setup guide.
↗Python SDK Integration Guide
SDK installation, authentication and examples↗TypeScript SDK Integration Guide
SDK installation, authentication and examples
↗
↗
↗This is only a basic call example. See the full docs for more parameters and advanced usage.
View Docs