| .. | ||
| README.md | ||
| txt2img_api.json | ||
ComfyUI workflows
txt2img_api.json uses only built-in ComfyUI nodes. It is a connection and
round-trip baseline, not the production sprite workflow. Set Studio's Model
field to the exact checkpoint filename installed on the worker.
Validate it locally:
sf gpu workflow-check examples\comfyui\txt2img_api.json
The production workflow should be exported from ComfyUI in API format and add:
- IP-Adapter/reference conditioning fed by
{{REFERENCE_0}}; - pose/depth ControlNet fed by
{{CONTROL_CONTROL}}; - an inpaint branch fed by
{{CONTROL_MASK}}; - low-denoise previous-frame img2img fed by
{{CONTROL_PREVIOUS}}; - alpha/background removal before
SaveImagewhere supported.
Keep the standard SpriteForge placeholders in the appropriate node inputs, then
run sf gpu workflow-check again.