prompt | string | yes | - | Text description for image generation. |
generation_image_size | enum | no | "square" | Output image dimensions (e.g. 384x384, 768x768). · Allowed: square, square_hd, portrait_4_3, portrait_16_9, landscape_4_3, landscape_16_9 |
generation_temperature | number | no | 1.0 | Sampling temperature for image generation. Higher = more variation. · Range: 0.1 – 2 |
generation_cfg_weight | number | no | 5 | Classifier-free guidance weight. Higher = closer to prompt but less creative. · Range: 1 – 20 |
generation_num_images | number | no | 1 | How many images to generate (1-4). · Range: 1 – 4 |
generation_seed | number | no | - | Random seed for reproducibility. Same seed + prompt = same output. |
image_url | string | no | - | URL of an image to analyze (image-understanding mode). |
question | string | no | - | Question to answer about the supplied image. |
analysis_temperature | number | no | 0.1 | Sampling temperature for image analysis. Higher = more variation. · Range: 0 – 1 |
analysis_top_p | number | no | 0.95 | Nucleus sampling probability mass for image analysis. · Range: 0.1 – 1 |
analysis_seed | number | no | - | Random seed for analysis output reproducibility. |