Kontext Image enlargement without distortion — v1.0

Use “Kontext Image enlargement without distortion” to enlarge an image while keeping a character’s appearance consistent. It uses kontext dev and Deblur LoRA, with a high-resolution enlargement as the intended output.

It combines kontext dev with Deblur LoRA, and its prompts maximize consistency of a character’s appearance.

The control can be too strong, so high-resolution processing may fail.

Input: an image. Intended output: a high-resolution enlargement with consistent character appearance.

Find it on Civitai at https://civitai.com/models/1886884. It is published by BBS Revenger and uses Flux.1 Kontext as its base model.

Node packs listed for the workflow: cg-use-everywhere, comfyui_faceanalysis, comfyui-custom-scripts, comfyui-easy-use, and rgthree-comfy.

Model files listed: 2xNomosUni_esrgan_multijpg.pth, ae.safetensors, clip_l.safetensors, Flux_kontext_deblurB.safetensors, and flux1-krea-dev.safetensors.

Additional model files listed: kontextNsfwFp8_kontextNsfwFp8-C.safetensors, OpenFLUX.1-vae.safetensors, SkinTextureRealisticDetailed.safetensors, t5xxl_fp16.safetensors, and t5xxl_fp8_e4m3fn_scaled.safetensors.

If high-resolution processing fails, adjust “Deblur lora strength” and “Flux Guidance”; increasing CFG may also be needed.

Sugestão · não verificado

It cannot restore old, damaged photos.

Before using it, verify that these files are available: Flux_kontext_deblurB.safetensors, kontextNsfwFp8_kontextNsfwFp8-C.safetensors, SkinTextureRealisticDetailed.safetensors, OpenFLUX.1-vae.safetensors, and 2xNomosUni_esrgan_multijpg.pth.

Sugestão · não verificado