Made with these models
The 17 models this integration runs.
Background removal (BiRefNet)
birefnetbackground-removal
esrgan on fal_ai
esrganimage-upscale
Fast, cheap drafts (FLUX.2 klein)
flux-2-klein-4btext-to-image
flux-dev on fal_ai
flux-devtext-to-image
Game objects on a plain ground
flux-game-assets-2dtext-to-image
Neutral-ground concepts
flux-game-assets-3dtext-to-image
flux-pro on fal_ai
flux-protext-to-image
flux-schnell on fal_ai
flux-schnelltext-to-image
Images with legible text
qwen-imagetext-to-image
Edit an image by instruction
qwen-image-edit-2511text-to-image
Cheap background removal (rembg)
rembgbackground-removal
sd-3.5 on fal_ai
sd-3.5text-to-image
sdxl on fal_ai
sdxltext-to-image
sdxl-controlnet-union on fal_ai
sdxl-controlnet-unionimage-to-image
Sharp upscaling (SeedVR2)
seedvr-upscaleimage-upscale
Edge-guided images (Z-Image)
z-image-turbo-controlnetimage-to-image
Seamless tiling textures
z-image-turbo-tilingtext-to-image
Fast, high-quality image generation. The default engine behind every new pipeline.
- Default image engine for a new workspace
- Text-to-image and image-to-image on one node
- Auto-switches to img2img when an image is connected
- Per-model step clamping so a graph never 422s on model swap
FLUX on FAL is the default image engine for a new Sushibelt workspace — it is what your first pipeline runs on. Attach it to a Diffusion Worker, wire a prompt in, and you have images.
It handles both text-to-image and image-to-image: connect an upstream image and the node switches to img2img automatically, so refinement passes need no reconfiguration. Aspect ratio, step count and guidance are inspector fields, and step counts are clamped per model so a graph authored for one model does not fail on another.
Outputs land in your library and can feed a 2D-to-3D node, a delivery node, or another diffusion pass.
imageimage-generationfluxtext-to-imageimage-to-imagefal