This workflow is designed to provide the easiest way to use the SDXL Refiner in ComfyUI.
It’s perfect for those who want to understand and learn the basics of how the Refiner works.
The nodes are neatly categorized into sections for better clarity,
and I’ve added memos to explain any complex features or functionality.
If you encounter any "missing node" errors while using this workflow, don’t worry!
Simply click Install Missing Node, and most issues should be resolved.
Additionally, make sure you keep your ComfyUI updated to the latest version for the best experience.
Feel free to explore and tweak this workflow to suit your needs.
I hope it helps you get started with SDXL Refiner in a straightforward and enjoyable way!
stable-diffusion-xl-refiner-1.0 Download link : https://huggingface.co/stabilityai/stable-diffusion-xl-refiner-1.0/tree/main
========================================================
ComfyUI에서 SDXL Refiner에 대한 워크플로우입니다.
Refiner의 기본적인 작동 원리와 사용법을 배우고 싶은 분들에게 적합합니다.
노드들은 섹션별로 깔끔하게 정리되어 있으며,
어려운 기능에 대해서는 메모를 통해 설명을 추가해 두었습니다.
만약 워크플로우를 사용할 때 "missing node" 오류가 발생한다면 걱정하지 마세요!
Install Missing Node 버튼을 누르면 대부분의 문제가 해결됩니다.
또한, ComfyUI를 최신 버전으로 유지하면 더욱 원활하게 사용할 수 있습니다.
이 워크플로우로 기본을 이해하고 필요에 따라 자유롭게 조정해 보세요!
stable-diffusion-xl-refiner-1.0 Download link : https://huggingface.co/stabilityai/stable-diffusion-xl-refiner-1.0/tree/main
========================================================
[contact]
Youtube : https://www.youtube.com/@A01demort
Instagram : https://www.instagram.com/a01demort/
Naver Cafe : https://cafe.naver.com/sdfkorea
Description
FAQ
Comments (3)
you answered alot of questions in those notes.. and setup was easy to follow and adapt
but still had to google
end_at_steps
found this
Addendum, the default "end_at_step 10000", is just a way of saying "Run until 'Steps'" since the start/end at values are clamped to the job itself.
@mystifying Thanks for Buzz
As I understand it, the reason the 'end_at_step' value is set to 10000 is as follows 'end_at_step' value of 10000 in ComfyUI's KSampler Advanced is simply an arbitrarily large number.
The actual sampling process will still stop at your initially set total steps (e.g., 30 steps), regardless of whether you set the end_at_step to 30, 100, or 10000.
The system will respect your total step count as the true endpoint of the process.
So, you can change it to match the step count.
However, since it’s tedious to adjust it every time, leaving it at 10000 doesn’t seem like a bad idea.
@a01demort totally with you.. also changed sampler and scheduler to inputs and use one spot to change both samplers