mood | string | Explicit-mode only: full mood prompt text. Ignored when agentMode='1'. |
size | string | Output aspect ratio. Defaults to 1024x1024. Ignored when agentMode='1' (picker chooses). |
image | string | Source image, multipart file upload. Either `image` (file) or `imageUrl` (string) must be present. |
scene | string | Explicit-mode only: full scene prompt text. Ignored when agentMode='1'. |
action | string | Explicit-mode only: full action prompt text. Ignored when agentMode='1'. |
intent | string | Free-text description of the desired vibe (e.g. 'noir detective vibe at midnight'). Optional; if omitted the picker decides purely from the source image. |
imageUrl | string | Alternative to `image`: publicly fetchable URL (HTTPS or IPFS) of the source image. |
agentMode | string | Set to '1' to opt into the server-side picker (recommended for autonomous agents). The picker chooses scene/action/mood/size from the curated catalog based on the source image plus the optional `intent`. |
sourceKind | string | What the source image actually is. 'photo' (default) runs the gpt-4o-mini describer to extract identity markers as text; source pixels never reach the renderer. 'gvc-token' skips the describer entirely and injects the source image as a Flux reference, preserving body color, character type (Robot / Default / Alien), hair, and accessories at the pixel level. |
sceneBgImages | string | Explicit-mode only: comma-separated list of scene background reference filenames (from the curated catalog). |