Have a ComfyUI workflow? Get a readable Workflow Doc in your browser.Make a workflow readable.Try it now ↗
⇩ComfyUI Workflow 🗎 | ⇩ InvokeAI Workflow 🗎
Runtime profile
Source description
⇩ComfyUI Workflow 🗎 | ⇩ InvokeAI Workflow 🗎
Tiled upscale workflows for ComfyUI and InvokeAI utilizing Flux.2 Klein models. The idea is similar to HiresFix from A1111. The image is first upscaled with ESRGAN model and then split into tiles and processed using the Klein model one by one. The tiles are then stitched back together and, optionally, color corrected. Since the original image is passed to Klein as reference, it helps preserve the image structure and style, while significantly reducing upscale artifacts and adds small amount of detail to the final image.
For best results correct errors and inpaint small details before upscaling.
Note: ComfyUI variant requires Steudio node pack .
AI-generated commentary
AI-generated explanation based on source and configuration details. Suggestions are clearly labeled.
This workflow is for enlarging images in ComfyUI or InvokeAI. It takes an image, upscales it with ESRGAN, processes tiled sections with Flux.2 Klein, and produces a stitched, upscaled image.
The image is first upscaled with ESRGAN, split into tiles, and processed one tile at a time with Flux.2 Klein.
The processed tiles are stitched back together, with optional color correction afterward.
The original image is passed to Flux.2 Klein as a reference to help preserve its structure and style.
Workflow files are available for ComfyUI and InvokeAI.
The ComfyUI variant requires the ComfyUI_Steudio node pack.
The listed required files are flux-2-klein-9b-fp8.safetensors, flux-2-vae.safetensors, full_encoder_small_decoder.safetensors, and qwen_3_8b_fp8mixed.safetensors.
full_encoder_small_decoder.safetensors is associated with black-forest-labs/FLUX.2-small-decoder.
Color correction can be applied after the tiles are combined.
Before upscaling, correct image errors and inpaint small details.
Suggestion · not verified
Before setup, verify which flux-2-vae.safetensors file to use; its checksum and provider source are not specified.
Suggestion · not verified
Does this need to be edited?
Sign in to send an edit request.
Sources
1 sourceSource excerpts
3 excerptsSource context: 84 downloads · Type Workflows · Base model Other
Estimated VRAM requirement
Estimate unavailable
16.9 GB across 2 of 4 model files. Model file total + 25% loading overhead + 2 GB execution buffer, rounded up.
Requirements
5 requirementsCheckpoint · 8.79 GB · SAFETENSORS · Unknown
flux-2-vae.safetensors
Not resolvedVAE · Unknown
full_encoder_small_decoder.safetensors
File unverifiedVAE · Hugging Face · black-forest-labs/FLUX.2-small-decoder · FLUX.2 Small Decoder
Text encoder · 8.07 GB · SAFETENSORS · fp16 · Unknown
Node pack · Registry
Source context: 83 downloads · Type Workflows · Base model Other
Source context: 75 downloads · Type Workflows · Base model Other
