Description #

The Seedream image generation model can generate high-quality images based on user input text or reference images, supporting multi-image grouping and streaming output.

Function overview #

• The Doubao Seedream series image creation model supports text, single-image, and multi-image input, enabling multi-image fusion creation based on subject consistency, image editing, multi-image generation, and other diverse features, making image creation more flexible and controllable. The newly released doubao-seedream-5.0-lite model offers enhanced generation quality in consistency, instruction adherence, text rendering, knowledge richness, and professional scenarios, better meeting enterprise-level visual creation needs. The model advantages are as follows:
  • Strong timeliness: doubao-seedream-5.0-lite is equipped with internet search functionality for the first time. The model autonomously determines whether to search internet content based on user prompts, integrates real-time online information, and improves the timeliness of image generation.
  • Strong instruction adherence: The alignment between generated images and input text is improved, reducing the illusion of “text-image mismatch” and accurately responding to detailed text instructions.
  • Strong consistency: Maintains a high degree of uniformity in character features, style, and details across images, ensuring completely stable and reliable generation results.
• LAS has integrated doubao-seedream-5.0-lite, doubao-seedream-4.5, doubao-seedream-4.0, and doubao-seedream-3.0-t2i models.

Application scenarios #

The Seedream series models can be widely applied in e-commerce, gaming, film and television, animation, and design industries, profoundly transforming traditional content production methods and significantly improving the efficiency of visual content production.

• E-commerce materials: virtual try-on, product real scene images, product selling point images, secondary editing of product images, and related scenarios.
• Film and television creation: character design, scene design, storyboard generation, and converting sketches to final artwork.
• Advertising and marketing: marketing posters, cover design, event banners, and related scenarios.
• Entertainment features: portrait photography, creative group shots, face swapping, style transfer, emoji creation, and related scenarios.
• Education and office: knowledge illustrations, graphic posters, popular science infographics, chart generation, and related scenarios.
• Game design: character design, scene design, prop design, UI design, and related scenarios.

Limitations #

• doubao-seedThe ream-5.0-lite model supports 2K and 3K resolution output.
• doubao-seedWithin the ream series, only the 4.0, 4.5, and 5.0-lite models support generating image sets. The number of input reference images plus the number of final generated images must not exceed 15.
• doubao-seedWithin the ream series, only the 4.0, 4.5, and 5.0-lite models support streaming output.
• doubao-seedThe ream-3.0-t2i model supports only text-to-image generation.

Image generation (non-streaming) #

POST https://operator.las.ap-southeast-1.volces.com/api/v1/online/images/generations

API description #

Invoke the Seedream image generation model to perform text-to-image, image-to-image, or image set generation (non-streaming).

Request parameters #

Response parameters #

Example #

Request example

curl --location "https://operator.las.ap-southeast-1.volces.com/api/v1/images/generations" \
--header "Content-Type: application/json" \
--header "Authorization: Bearer $LAS_API_KEY" \
--data '{
    "model": "doubao-seedream-4-5-251128",
    "prompt": "",
    "size": "2048x2048",
    "response_format": "url",
    "watermark": true
}'

Response example

{
    "model": "doubao-seedream-4-5-251128",
    "created": 1764041608,
    "data": [
        {
            "url": "<image_url>",
            "size": "2048x2048"
        }
    ],
    "usage": {
        "generated_images": 1,
        "output_tokens": 16384,
        "total_tokens": 16384
    }
}

Image generation (streaming) #

POST https://operator.las.ap-southeast-1.volces.com/api/v1/online/images/generations

API description #

Invoke the Seedream image generation model to perform text-to-image, image-to-image, or image set generation (streaming SSE).

Request parameters #

Response parameters #

Example #

Request example

curl --location "https://operator.las.ap-southeast-1.volces.com/api/v1/images/generations" \
--header "Content-Type: application/json" \
--header "Authorization: Bearer $LAS_API_KEY" \
--data '{
    "model": "doubao-seedream-4-5-251128",
    "prompt": "",
    "size": "2048x2048",
    "response_format": "url",
    "stream": true
}'

Response example

{
    "type": "image_generation.completed",
    "model": "doubao-seedream-4-5-251128",
    "created": 1589478378,
    "usage": {
        "generated_images": 1,
        "output_tokens": 16384,
        "total_tokens": 16384
    }
}