Instant Model Boost with SDXL FreeU - ComfyUI: No Extra Time, No Extra Cost!
Вставка
- Опубліковано 6 жов 2024
- In this tutorial, I'm exploring SDXL FreeU through ComfyUI, sharing how it can upgrade models without extra time or costs.
Check the project here: github.com/Che... and see how I've integrated it into my workflow: civitai.com/mo....
This is a new venture for me , and I look forward to learning and sharing more AI magic .
Good to see more tutorials on here and best of luck going forward! Just an observation, I struggled to see some of the settings on the nodes. Maybe increase from 720 to 1080?. Then lastly I am playing with FreeU, however i still need to see a video or resource that excplain exactly what the parameters do in a easy to understand way. I am forever greatful for the developers that give us these great resources but I feel they forget that our mere mortals dont understand the jargon
Note, I've adjusted that for the next one. Crashing my computer in the process but will figure out something to solve that for the next ones ;)
Good content BUT 1080 is a must nowadays, it's too small to see/read
Noted! I'll do this on the next one, and hopefully without blowing up my GPU in the process. :)
Thank you for pointing this node out and sharing some settings. I'd like to know more about what those settings each do in freeu
Your no 13 sub. Keep it up. I would hope in the future we get a method for Automatic SDXL
Easy Peasy 🙂 useful worklflow, thanks man
Great video!! Many thanks for sharing your knowledge
super helpful on many fronts. thank you!
Good job ! Love so much comfy ui, I hope more tuto on workflow built :) (btw can you tell me please what is the green and red button inside your primitive boxes ?)
A new one is up! You can find the green and red box in there ;) just change pink hair to beige hair. ;)
I would suggest using a separate browser for ComfyUI.
I can't click on the links in Description. Please edit them. Thank you very much for sharing.
Thanks for pointing that out. It seems I need to verify my account first in order to post external links. Verification is on the way , links should work within 24H.
civitai.com/models/151390/freeu-sdxl-simple || github.com/ChenyangSi/FreeU
@@tech_bytebrain Nice. Thank you again.
Amazing! Can you do Animatediff animation with consistency?
Will cover fundamentals on text to image > img to img (Ctrlnet + IPAdapter) than we'll cover video and more complex things ;)
👋
Thanked youuuu!
What are your FreeU settings? Is there any guides? Basically it's useless if you don't know how every parameter works. However, leaving every setting to 1.0 often produces better result than without FreeU. And if your image has faults like bad fingers or pose, then setting b1 or b2 to 0.8, (but not both at the same time), tends to fix the image. 0.8 is like a sweetpot there.
EDIT: Your settings didn't work for me. Your settings fried my images and gave them washed out look. My settings probably don't work for you either. It depends on art style and model that you use, cfg setting, prompt etc.. Basically FreeU is just another tool in your bag, not the decisive answer to everything. I'm using Fooocus-MRE.
What is the green and red icons inside your positive prompt when you paste the prompt? Is it a plugin in comfyui?
i think thats grammarly, it's a browser extension that gives you feedback on your grammar as you type
Correct!
Just use batch 3 bro, wasting of space
please SDXL + controlnet
Sure, any specifics? Or, just general Controlnet (bit of everything)? .
@@tech_bytebrain a bit of everything please, to start :)
@@tech_bytebrain also, IPAdapter 😉
How about:
2 cnet's x/y/z in different weights
3 cnet's x/y/z in different weights
A workflow where we can see the result of a prompt in different controlnets.
@@ouzyclYes interesting