Seedream 5.5 is on fal with edit and layer endpoints

fal added Seedream 5.5, ByteDance's image model, with separate endpoints for generating an image, editing one and splitting one into layers.

The edit endpoint takes a text instruction and up to 10 reference images, and fal lists multi-turn refinement and editing text inside an image among the things the model does. Layerize breaks an image into a base image plus independently editable PNG layers, returning each layer's position and stacking order, with optional text recognition. Text-to-image and edit can both return a transparent PNG.

fal bills per output image, at $0.03 for 1K or 1.5K and $0.0405 for 2K. Edit adds $0.012 per input image for each output. Layerize counts the base image and every separated layer as an output, and adds $0.012 once per request for the input image.

The endpoint IDs are bytedance/seedream/v5.5/text-to-image, bytedance/seedream/v5.5/edit and bytedance/seedream/v5.5/layerize.