PixelLab pixel-art generation API — text-to-image (Pixflux), style transfer (Bitforge), animation (skeleton + text), rotation, inpainting, skeleton estimation. Images flow through the package filespace: inputs are paths to files (usually outputs of earlier calls), outputs are written as PNG files and returned as paths.
| Command |
Kind |
Description |
| generate pixflux |
action |
Text-to-image generation (Pixflux model). Width/height 16-400 each, area between 32x32 and 400x400 (e.g. 64x64, 128x96). Optional style enums (outline, shading, detail, view, direction) are weakly guiding; init_image/color_image are filespace paths; seed for reproducibility. Output: one PNG file (flat name: pixflux-<timestamp>_s<seed>.png, or your filename). |
| generate bitforge |
action |
Style-transfer generation (Bitforge model): generate from a description while transferring the style of a style_image. Width/height 16-200 each, area at most 200x200. Supports inpainting (inpainting_image + mask_image), forced palette, transparent background, skeleton guidance. Output: one PNG file (flat name: bitforge-<timestamp>_s<seed>.png, or your filename). |
| animate skeleton |
action |
Generate 4-frame animation windows from skeleton poses (Animate with skeleton). Width and height each from 16/32/64/128/256. skeleton_keypoints: JSON array of EXACTLY 3 frames of keypoints (as returned by 'skeleton estimate') — the model is a 3-frame window; pass consecutive pose frames. reference_image is the character on a transparent background (filespace path). Output: 4 PNG files (flat names <base>_f1..f4). |
| animate text |
action |
Text-guided animation: describe a character and an action, get 4 animation frames. Size is fixed at 64x64 (width/height optional, must be 64). Longer animations via start_frame_index windows (model always generates 4 frames). reference_image required (filespace path). Output: 4 PNG files (flat names <base>_f1..f4). |
| rotate |
action |
Rotate a character or object: turn it between camera views (from_view/to_view) and directions (from_direction/to_direction), or by degrees (view_change/direction_change). Width/height each from 16/32/64/128. from_image required (filespace path); mask_image (optional) requires init_image. Output: one PNG file (flat name: rotate-<timestamp>_s<seed>.png, or your filename). |
| inpaint |
action |
Edit existing pixel art: describe the change; white regions of mask_image are regenerated on inpainting_image. Width/height 16-200 each, area at most 200x200. Both inpainting_image and mask_image required (filespace paths). Output: one PNG file (flat name: inpaint-<timestamp>_s<seed>.png, or your filename). |
| skeleton estimate |
action |
Estimate the skeleton of a character image: returns 18 labeled keypoints (with z_index for depth ordering) to edit and feed back into 'animate skeleton'. Input image is a filespace path (character on a transparent background); size 16-256 per side. No image output — keypoints come back as data. |
| balance get |
action |
Current PixelLab account balance in USD. Cheap pre-flight before generation runs. |