← sources

Krea 2 General Edit V2 – what results are people actually getting?

source: reddit (reddit:comfyui)published 2026-11-03 10:51ingested 2026-11-04 02:41

retained text

I've been experimenting with Krea 2 image editing in ComfyUI, particularly the new General Edit V2 LoRA, and I'm interested in hearing about other people's experiences.

Krea 2 produces excellent Text To Image results, but so far I've struggled to achieve anything approaching that same level of quality and consistency when editing existing images.

I'm using the ComfyUI-Krea2Edit custom nodes, but the published example workflow predates the V2 LoRAs. I've therefore also raised an issue with the developer asking whether the original workflow is still recommended and what settings are appropriate for General Edit V2.

A few things I'd particularly like to understand:

* Are you getting consistently good results with General Edit V2?
* Are you using the original sample workflow, or have you made significant changes to it?
* Do you find Turbo or Raw gives better editing results?
* What LoRA strength, sampling steps, CFG and reference boost settings work best?
* Does reducing reference boost help when the model is reluctant to make the requested change?
* Are there particular types of editing where Krea 2 performs especially well, or struggles?

I'm particularly interested in ordinary instruction-based edits: enhancing/reskinning images, removing or replacing objects, altering backgrounds, relighting and making local changes while retaining the rest of the original image.

I'm not expecting every edit to work perfectly, but I'd like to establish whether my disappointing results are down to workflow configuration, the limitations of the current Edit LoRAs, or some combination of the two.

I'd appreciate hearing about people's actual experience, including settings and workflow examples if possible. Thanks!

document fetched 2026-11-04 00:41 · canonical https://reddit.com/r/comfyui/comments/1xmpvcq/krea_2_general_edit_v2_what_results_are_people