> ## Documentation Index > Fetch the complete documentation index at: https://docs.anyfast.ai/llms.txt > Use this file to discover all available pages before exploring further. # Seedream 5.0 Flash > Generate, edit, combine, preserve transparency, and decompose images into editable layers with Seedream 5.0 Flash. Seedream 5.0 Flash is ByteDance's faster, lower-cost model for image generation and editing. Use model ID `doubao-seedream-5-0-flash-260915` with AnyFast's `POST /v1/images/generations` endpoint. The upstream model supports text-to-image generation, single- and multi-image editing, transparent-background editing, and layer decomposition. Seedream 5.0 Flash generates one image per normal generation request. It does not support grouped image generation, web search, or streaming. Do not send `sequential_image_generation`, `sequential_image_generation_options`, `tools`, or `stream`. ## Key capabilities * **Text to image** — Generate one image from a text prompt * **Image editing and fusion** — Edit one reference or combine content and style from up to 10 references * **Interactive editing** — Point to regions with natural language, markers, points, bounding boxes, arrows, or sketches * **Transparent background** — Preserve an alpha channel in a single-image editing workflow * **Layer decomposition** — Split one image into a background plus up to 16 editable transparent PNG layers * **Multilingual text** — In addition to Chinese and English, supports Russian, Arabic, Filipino, Thai, Turkish, Korean, Malay, Spanish, Portuguese, Indonesian, French, German, Vietnamese, and Japanese prompts ## Generate from text ```bash cURL theme={null} curl https://www.anyfast.ai/v1/images/generations \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "doubao-seedream-5-0-flash-260915", "prompt": "A premium product photograph of a white ceramic cup on a light gray background, centered composition, soft shadow, no text", "size": "1.5K", "response_format": "url", "output_format": "jpeg", "watermark": false }' ``` ```python Python theme={null} import requests response = requests.post( "https://www.anyfast.ai/v1/images/generations", headers={ "Authorization": "Bearer YOUR_API_KEY", "Content-Type": "application/json", }, json={ "model": "doubao-seedream-5-0-flash-260915", "prompt": "A premium product photograph of a white ceramic cup on a light gray background, centered composition, soft shadow, no text", "size": "1.5K", "response_format": "url", "output_format": "jpeg", "watermark": False, }, ) print(response.json()) ``` `size` accepts `1K`, `1.5K`, `2K`, or a valid explicit `widthxheight` value. Flash supports only `standard` prompt optimization; do not send `optimize_prompt_options.mode: "fast"`. ## Edit or combine references ```bash cURL theme={null} curl https://www.anyfast.ai/v1/images/generations \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "doubao-seedream-5-0-flash-260915", "prompt": "Use the composition of image 1 and the visual style of image 2.", "image": [ "https://example.com/composition.png", "https://example.com/style.png" ], "size": "1K", "output_format": "jpeg", "watermark": false }' ``` `image` accepts a URL, a Base64 data URI, or an array of up to 10 references. In the prompt, refer to images by their array order, such as “image 1” and “image 2.” For coordinate-based editing, use coordinates in a 1000 × 1000 normalized coordinate system: ```json Coordinate-based edit theme={null} { "model": "doubao-seedream-5-0-flash-260915", "prompt": "Place the subject from image 1 179 283 796 986 into image 2 118 331 933 871.", "image": [ "https://example.com/source.png", "https://example.com/target.png" ], "size": "2K", "output_format": "png" } ``` ## Preserve transparency ```bash cURL theme={null} curl https://www.anyfast.ai/v1/images/generations \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "doubao-seedream-5-0-flash-260915", "prompt": "Keep the subject and refine the edges while preserving the transparent background.", "image": "https://example.com/transparent-source.png", "background": "transparent", "size": "1K", "output_format": "png", "watermark": false }' ``` Transparent mode is available only for image-to-image requests with exactly one input image that already has an alpha channel. Use `output_format: "png"`; JPEG input or output is not compatible with transparency. ## Split an image into layers ```bash cURL theme={null} curl https://www.anyfast.ai/v1/images/generations \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "doubao-seedream-5-0-flash-260915", "prompt": "Separate the green leaf icon from the white background into one editable layer.", "image": "https://example.com/green-leaf.jpeg", "layer_decomposition": true, "size": "1K", "response_format": "url", "output_format": "jpeg", "watermark": false }' ``` Layer decomposition accepts exactly one PNG or JPEG. It returns one background image and up to 16 transparent PNG layers. The background follows `output_format`; every extracted layer is PNG and includes `z_index`, `bounding_box`, `name`, and `description` metadata. ```json Response shape theme={null} { "model": "dola-seedream-5-0-flash-260915", "created": 1792129605, "data": [ { "url": "https://example.com/seedream-flash-layer-0.jpeg", "size": "1024x1024", "output_format": "jpeg", "z_index": 0 }, { "url": "https://example.com/seedream-flash-layer-1.png", "size": "662x855", "output_format": "png", "z_index": 1, "bounding_box": { "absolute": [ 305, 251, 729, 800 ], "normalized": [ 298, 245, 711, 780 ] }, "name": "green leaf icon", "description": "Extract the complete green leaf icon with leaf blade, veins and petiole, no other background elements included" } ], "usage": { "input_images": 1, "generated_images": 2, "output_tokens": 8549, "total_tokens": 8549 } } ``` For decomposition, `size` accepts `auto`, `1K`, `1.5K`, or `2K`; explicit pixel sizes are not supported. If any requested layer fails, the whole request fails. ## Output controls | Parameter | Values | Notes | | - | - | - | | `size` | `1K`, `1.5K`, `2K`, or explicit pixels | Normal generation defaults to `2K`; layer decomposition also supports `auto` | | `response_format` | `url`, `b64_json` | URLs are valid for 24 hours; save them promptly | | `output_format` | `jpeg`, `png` | Defaults to `jpeg`; transparent output and extracted layers use PNG | | `background` | `opaque`, `transparent` | `transparent` is restricted to the single transparent-input workflow | | `watermark` | `true`, `false` | Defaults to `true` | | `optimize_prompt_options.mode` | `standard` | Default. Flash does not support `fast` mode. | ## Size and input limits For normal generation, explicit dimensions must contain between 921,600 and 4,624,220 total pixels and use an aspect ratio in `[1/16, 16]`. Representative sizes include `1024x1024` for 1K, `1536x1536` for 1.5K, and `2048x2048` for 2K. Normal reference images support `jpeg`, `png`, `webp`, `bmp`, `tiff`, `gif`, `heic`, and `heif`. Each image must be smaller than 30 MB, have an aspect ratio in `[1/16, 16]`, have width and height greater than 14 px, and contain from 196 through 36,000,000 pixels. Layer decomposition accepts one PNG or JPEG with 512×512 to 6000×6000 total pixels, with the same aspect-ratio and 30 MB limits. View all Seedream 5.0 Flash workflows. ## Official references * [Seedream 5.0 Pro / Flash guide](https://docs.volcengine.com/docs/ark/seedream-5-0-pro?lang=zh) * [Image generation API](https://docs.volcengine.com/docs/ark/image-generation-api?lang=zh) * [Seedream model comparison](https://console.volcengine.com/ark/region:cn-beijing/docs/seedream-4-0-5-0)