Image Generation

Model name, fixed value: jimeng_t2i_v40

Service representation, fixed value: jimeng_t2i_v40

Prompts for image generation support both Chinese and English input. Recommendations: The maximum length should not exceed 800 characters, as overly long prompts may result in abnormal image generation or failure to take effect. It supports directly specifying the image aspect ratio in the prompt, and the model will intelligently determine the aspect ratio based on the "size" field, generating 2K resolution images with an aspect ratio not exceeding 3 by default. It is not recommended to input special symbols such as $ other than "".

Area of the generated image, default value: 4194304, i.e., 20482048, generating a 2K resolution image; Value range: [10241024, 40964096], capable of generating 1K to 4K resolution images; It is recommended to generate images above 2K resolution, as images with too low a resolution are highly likely to have issues such as poor facial rendering and abnormal text; if the output of a 2K image does not meet expectations, try outputting a higher resolution image. Text-to-image scenario: Either area or width and height must be provided; if neither is provided, the default area is 20482048, generating a 2K resolution image with the model intelligently determining the aspect ratio; when both area and width and height are provided, width and height take priority; if only area is provided, the model will intelligently determine the aspect ratio based on the user's prompt intent. Image-to-image scenario: The model intelligently determines the aspect ratio based on the user's prompt intent and the reference image size.

Width and height of the generated image; this field is not passed by default. The product of width and height must be within [10241024, 40964096], and the aspect ratio must be within [min_ratio, max_ratio]. Note: Both width and height must be passed for this to take effect. Recommended width and height options: 1K: 1024x1024 (1:1); 2K: 2048x2048 (1:1), 2304x1728 (4:3), 2496x1664 (3:2), 2560x1440 (16:9), 3024x1296 (21:9); 4K: 4096x4096 (1:1), 4694x3520 (4:3), 4992x3328 (3:2), 5404x3040 (16:9), 6198x2656 (21:9).

Width and height of the generated image; this field is omitted by default. The product of width and height must be within [10241024, 40964096], and the aspect ratio must be within [min_ratio, max_ratio]. Note: Both width and height must be passed for this to take effect.

The degree of influence of the text description; a larger value indicates a greater influence of the text description and a lesser influence of the input image (precision: supports two decimal places). Default value: 0.5. Value range: [0, 1].

Whether to force single image generation. Default value: false. Note: The more images generated in a set, the longer the processing time, and the generation time will increase significantly as the number of images grows; for stable multi-image output, it is recommended to use the prompt to limit the number of images to 9 or fewer; when generating multiple images, it is recommended to explicitly specify the image resolution in the prompt or directly pass specific width and height parameters to avoid API errors caused by inconsistent resolutions among the generated images.

The width/height ratio of the generated image ≥ min_ratio; if the smart aspect ratio < min_ratio, min_ratio will be used. Default: 1/3. Range: [1/16, 16).

The width/height of the generated image ≤ max_ratio; if the smart aspect ratio > max_ratio, max_ratio will be used. Default value: 3. Value range: [1/16, 16)

When using asynchronous callbacks, the callback endpoint URL (must be publicly accessible)

When using asynchronous callbacks, configure whether to return image resources as URLs (valid for 24 hours). Default value: false

Set watermark information when using asynchronous callbacks

Set the implicit identifier when using asynchronous callbacks.