UPDATED TO VERSION 2
VERSION 2 is a complete upgrade to quality and resolution. utlizing the FULL step count of both models with a small upscaler "realESERGAN2x" on the FIRST pass to ensure that the second pass of Z-Image is handed the latent data at 2048x2048 for use of full resolution capabilities of the model. this gives the final result a large resolution for minimal vram requirement and a highly detailed and refined image.
2 pass image workflow for detail refinement.
1st Model - Kandinsky Image Lite (base image)
2nd Model - Z-Image BASE gguf (detail refiner)
essentially it goes:
Kandinsky generates at 1024x1024 (maximum). you generate at 100 steps and then into the upscaler to bring the resolution to 2048x2048 (maximum). this passes the latent data to Z-Image to run 50 steps at the now larger resolution.
Z-Image base can NATIVELY GENERATE AT 2048X2048!
(this means after the low grade upscale, theres no artifacting because Z-Image corrects it in the second pass!)
REQUIRES KJNODES
REQUIRES realESERGAN2x
REQUIRES UNLOAD MODELS NODE
ALL CAN BE FOUND IN THE COMFY MANAGER
note: generation times will increase significantly. this is not about speed, this is about QUALITY.
__________________________________________________________________________________________________
2 pass image workflow for detail refinement.
1st Model - Kandinsky Image Lite (base image)
2nd Model - Z-Image BASE gguf (detail refiner)
VERSION 1
the kandinsky model acts as your first pass model, generating 20-30 steps to create an overall composition. the z-image base model runs on the second pass as a detail refiner, running 20-30 steps. this when playing with the denoise on the SECOND PASS can provide some interesting results!
Supported Resolutions
Square: 1024 × 1024
Landscape: 1280 × 768
Portrait: 768 × 1280
NOTE: i chose to run the gguf model for the zimage base. this workflow is vram intense and using the gguf takes some of the pressure off. the regular load diffusion model node is in the workflow and can be unbypassed and strung in if youd like to try using the full weight model instead!
CUSTOM NODE REQUIREMENTS:
Kiko Tools Nodes
Easy Use Nodes
Unload Model Nodes
the unload model node is NECESSARY to prevent OOM. all custom nodes can be found in the comfyui custom nodes manager!
Kandinsky files can be downloaded in comfyui templates!