Assetgenfile Format DocumentationASSET) must appear before any ASSET entries.PROMPT lines; each line will be concatenated in order to form the full prompt.MODEL, SIZE, etc.) may appear only once.ASSET lines must come after every global directive. Each one defines a filename and its specific prompt.ASSET line is mandatory.PROMPT lines are optional but strongly recommended—they establish shared context for all assets and help ensure reproducible results.Global Directive Defaults If a global directive is omitted, it is simply not sent to the OpenAI API, and the API’s own default behavior applies. CLI defaults (when you do set them) are:
| Directive | Default Value | Notes |
|---|---|---|
MODEL |
dall-e-2 |
|
SIZE |
(API default) | DALL·E-2: 256x256gpt-image-1: auto |
BACKGROUND |
auto |
(gpt-image-1 only) |
MODERATION |
auto |
(gpt-image-1 only) |
OUTPUT_COMPRESSION |
100 |
(gpt-image-1 only; applies to JPEG/WEBP) |
QUALITY |
(API default) | gpt-image-1: autodall-e-3: hddall-e-2: standard |
STYLE |
vivid |
(dall-e-3 only) |
ASSET line in the form:
ASSET <filename> <asset-specific prompt>
PROMPT lines (in order) followed by the asset-specific prompt text..png, .jpeg, .webp) is used to auto-detect the output format.ASSET line without a valid filename or prompt will trigger a syntax error.MODELChoose which OpenAI image-generation model to use.
dall-e-2, dall-e-3, gpt-image-1gpt-image-1 for highest quality (most expensive)dall-e-3 for strong balance of quality & costdall-e-2 for basic assets or cost-sensitive needs256x256, 512x512, 1024x10241024x1024, 1792x1024, 1024x17921024x1024, 1536x1024, 1024x1536, autostandardhd, standardauto, high, medium, lowvivid, natural)SIZESpecifies image dimensions (maps to the API’s size parameter).
256x256, 512x512, 1024x10241024x1024, 1792x1024, 1024x17921024x1024, 1536x1024, 1024x1536, auto256x256 for DALL·E-2)auto (GPT-Image-1) lets the API chooseBACKGROUNDControls transparency (GPT-Image-1 only).
transparent, opaque, autoautotransparent for compositing, opaque for full scenesMODERATIONSets content moderation level (GPT-Image-1 only).
low, autoautolow for benign game assets, auto for general safetyOUTPUT_COMPRESSIONJPEG/WEBP compression level (GPT-Image-1 only).
0–100100QUALITYRendering fidelity.
standardhd, standardauto, high, medium, lowSTYLEStylistic bias (DALL·E-3 only).
vivid, naturalvividvivid for cartoons, natural for realismPROMPT A clean top-down 2-D sprite on a transparent background.
PROMPT Include subtle shadows for depth.
MODEL gpt-image-1
SIZE 1024x1024
BACKGROUND transparent
MODERATION low
OUTPUT_COMPRESSION 80
QUALITY high
ASSET road_straight_ns.png A seamless 256×256 asphalt road segment running north–south.
ASSET tree.png A stylized pine tree with dark green needles.
ASSET icon.webp A glowing energy orb with electric sparks.