Run on Badgr

ComfyUI

Run ComfyUI image and video workflows — SDXL, Flux, ControlNet, and custom node graphs — on GPU capacity with max-cost protection.

Run Comfy workflow Get help running this

Specific ComfyUI jobs

Generate images with Flux

Run Flux.1 text-to-image workflows in ComfyUI — high-fidelity generations from a text prompt, with optional LoRA and upscale nodes chained in.

Generate images with SDXL

Run Stable Diffusion XL text-to-image or image-to-image workflows in ComfyUI, with refiner pass and standard SDXL samplers.

Batch-generate images from a prompt list

Run one ComfyUI workflow across a list of prompts (CSV or JSONL) in a single job, producing one image set per prompt without re-launching per prompt.

Generate product photos on a clean background

Run a ComfyUI workflow tuned for e-commerce product shots — subject isolation, background removal/replacement, and consistent lighting across a product line.

Generate character reference sheets

Run a ComfyUI workflow that produces a consistent multi-pose/multi-angle character sheet from a reference image or LoRA, useful for game and illustration pipelines.

Upscale images

Run a ComfyUI upscale workflow (ESRGAN/tiled diffusion upscalers) to increase resolution on existing images without re-generating from scratch.

Inpaint or edit part of an image

Run a ComfyUI inpainting workflow to regenerate a masked region of an image — object removal, replacement, or targeted edits — while keeping the rest of the image untouched.

Generate images with ControlNet guidance

Run a ComfyUI workflow that uses ControlNet (pose, depth, canny, or lineart) to constrain generation to a reference structure while a prompt drives style and content.

Compatible GPU routes

L40S 48GB A40 48GB RTX 5090 32GB RTX 4090 24GB A6000 48GB

Example command

badgr run "comfyui --workflow workflow.json" \
  --gpu L40S \
  --max-cost 5

Common failure modes

Out of VRAM loading a large checkpoint or ControlNet stack

Move to a higher-VRAM route (L40S 48GB or A40 48GB) or enable tiled VAE/CPU offload in your workflow.

Workflow hangs on custom node install at startup

Pin custom node versions in your workflow JSON and use --max-runtime so a stuck install doesn't burn budget.

Slow queue times on shared budget GPUs

Check live route

View details →