3 Amazing and Fun Updates - ComfyUI and A1111 WebUI!
Vložit
- čas přidán 25. 05. 2024
- Have you ever wanted to have something positioned nicely in your Stable Diffusion AI art generations? Just like with the old Latent Coupling extension, we now have a nice way to add things to images without inpainting. Want two people side-by-side? No problem! Need a rodent wearing armour on the right? Not an issue. Central circle only? Thanks circle mask, that’s just what I wanted.
And that's not all! There is also a new way to create higher resolution images called “DeepShrink”, and a new IPAdapter model to boot.
Wow! :)
Update: A1111 extension now work in sd1.5 too! 😃
Loads of ComfyUI workflows & resource links at this URL -
github.com/nerdyrodent/AVeryC...
Stable Diffusion - Face + Pose + Clothing - NO training required!
• Stable Diffusion - Fac...
== More Stable Diffusion Stuff! ==
* How do I create an animated SD avatar? - • Create your own animat...
* Installing Anaconda for MS Windows Beginners - • Anaconda - Python Inst...
* Dreambooth Playlist - • Stable Diffusion Dream...
* Textual Inversion Playlist - • Stable Diffusion Textu...
Contents:
0:00 Introduction to new stable diffusion features
0:28 DeepShrink HiRes Fix
2:50 IPAdapter Attention Masks
6:21 IPAdapter full face - Věda a technologie
I really think you make the best videos. They're always some blend of more concise, thorough or more freah idea than many of the other channels, and you seem to embrace the concept of freely sharing knowledge as a core principle. Keep it up Mr Rodent 😂
You have a great niche. The nerdy rodent video on a topic is usually the best!
Not4Talent is good but infrequent, usually a very specific workflow, and less general.
Olivio is getting a bit annoying these days.
Sebastian Kamph is even slower, is hit and miss and uses clickbait.
SiliconThaumaturgy is unrivalled in depth but very infrequent.
AItrepreneur used to be ok, but less so recently.
Did I miss any?
Love the Reposer workflow. We have similar mindset. Good to go, ship it 😅👏🏽
New Nerdy Node! Non-noodly node. Neat notebook.
Narley.
Love your channel mate!
Always insightful - thanks!
Glad it was helpful!
Thank Mr. Rodent! 😊👍
2:11 you can edit the a1111 ui-config and change that to whatever size you want.
this is the way
Yup! Advanced users can edit the text file ;)
Can you also tell us the link to the new high res fix for a1111 :-) ?
I made a mask for 3 subjects or 2
subjects and the background and it works well
Thank you for the video !
Fantastic great video 👍
Thanks for sharing. Dear expert, if I want to change the file name of outputs from ComfyUI to a format like 'the xxth image - seed code', similar to the format used by SD outputs (00001-SEED), how should I set this up in ComfyUI?
But what about A1111 with the masking...?!
@nerdy rodent Quck question: What's the advantage of the new ip adapter face full compared to the old one.. in practice, it seems exactly the same ? Oh and thank you for showing me ip adapter in your earlier vids, i use is constantly. Also a neat trick is to use reactor on top of ip-adapter face, helps A TON with likeness. Just use the same face image for both .. also low cfg like 3.5 helps too !
It’s smaller and has more face in it 😉? Yeah, it’s pretty similar but I think it gives slightly more accurate output.
@NerdyRodent thx for the reply ! Yeah its much more accurate after using it for sure!
офигенно полезное видео, я очень рад
I love your videos but could you be more specific with the links in the description? I'm looking for the ComfyUI workflow image for the hires fix (from the beginning of the video) and I can't find it in your page
The hires fix is just a node you can plug in via “search”
Is it possible to add a positive and a negative prompt for each mask to guide the generation more precisely?
Would it work with two checkpoints for each side?
I think I love you.
Will you make the workflows available on your github ?
Sure, I’ll pop it up if you like. It’s not changed much from the example provided
Thank you very much @@NerdyRodent for putting it on your github ! It does save me quite some time ! I do appreciate it a lot ! Thanks ! :)
@@jimdelsol1941 did you get it to work? It works for me but does not add the images I upload. Just runs with out adding the images I provide.
this is very neat. Now can you do half-woman-half-rodent creature with this?
For sure 😉
Question: why not modify UI text file to increase size of generated image ?
That’s certainly something advanced users can indeed do!
with comfy how do we know the version ?
The face full doesn’t work properly with img2img sequencing…..bummer….😣
I just want to be able to organize things better, even with just 5 nodes it gets super messy… have a look at nuke/fusion or some other node based editor… take some queues from those.
trying to put me and my cat in the attention masking example, but struggling to keep my and my cat's faces accurate
is attention masking that much better than just inpainting?
I’d say give both a go and compare to see which you prefer from both a workflow and output perspective!
@@NerdyRodent Whoa you responded! Looks like I'll be giving this and comfyui a go in a few days 👌
Thank you for putting out such quality videos!
@@lechefski and thank you for watching! Hope you enjoy the things 😀
Can you please post the JSON links for the workflows you show in your video? I don't see them on the github link. Thanks.
👋
👋
I was today days old when I realized I was attracted to mousegirls
😆
In fact, Kohya's hiresfix increase the time needed.
With the regular hiresfix, you don't have to do 20 steps then 20 hires steps. If you want to do 20 steps, you can do 10 steps, upscale with anything but Latent, then 10 steps again. So, if your hires scale is 2, your image has 4 times more pixels. So kohya increase time required for the image render by about 60%.
But the images are also different, the composition are different, so... It's just a different workflow.
Ah, I only did a quick test of Koyha 1024x2048 (3seconds) Vs “normal” 512x1024 then 2x upscale, denoise 0.5 (5seconds, needed fixing more too) so figured it was faster as koyha decreased time needed by around 40%. Not to mention all the fiddling with the denoise to try and get it right ;) But yes, it’s another tool in the toolbox even if the speed is difficult on differing hardware!
thats not what i noticed...for me it chopped it from 9sec to 6sec
First!! nope, second, nope, third it is!!
I am not convinced about the new high res fix. I generated a 512x904 with 2x high res fix in 50 sec. After that I disabled the high res fix, set the resolution to 1024x1808 and enabled the Kohya High res fix. Sadly the time went up to 61 seconds
It’s just an option - the choice is yours!
It's also not working for me. It's taking significantly longer, at least 4x longer. The repo was updated right before I installed it though so I wonder if something go broken.
link?
@@Darkwing8707
i'm also not convinced that the Kohya hires is much different than the standard hires fix as far as quality...i thought it was at first, but after close comparisons, it seems like the highres fix settings gives better results, especially when upsampling.
SECOND!
Are 1.5 users going to be left in the dust? :(
No more toys for my 3gb vram sadge :( and I jus started having fun with ipadapter
That's one BIG RAT
Totally normal where I come from 😉
can I have your workflow please ?
Sure, workflows are at github.com/nerdyrodent/AVeryComfyNerd 😀
FIRST!
second :P
🫶
🩵
Why would anyone use this convoluted masking process over, say, inpainting?
Maybe update the title with the correct one, i.e. include "SDXL" in this. otherwise it's just another clickbait.
Clickbait is having something in the title that isn't true. The title is still true. Learn what clcikbait means before commenting next time.
°∆ I believe we are meant to be like Jesus in our hearts and not in our flesh. But be careful of AI, for it is just our flesh and that is it. It knows only things of the flesh (our fleshly desires) and cannot comprehend things of the spirit such as peace of heart (which comes from obeying God's Word). Whereas we are a spirit and we have a soul but live in the body (in the flesh). When you go to bed it is your flesh that sleeps but your spirit never sleeps (otherwise you have died physically) that is why you have dreams. More so, true love that endures and last is a thing of the heart (when I say 'heart', I mean 'spirit'). But fake love, pretentious love, love with expectations, love for classic reasons, love for material reasons and love for selfish reasons that is a thing of our flesh. In the beginning God said let us make man in our own image, according to our likeness. Take note, God is Spirit and God is Love. As Love He is the source of it. We also know that God is Omnipotent, for He creates out of nothing and He has no beginning and has no end. That means, our love is but a shadow of God's Love. True love looks around to see who is in need of your help, your smile, your possessions, your money, your strength, your quality time. Love forgives and forgets. Love wants for others what it wants for itself. Take note, true love works in conjunction with other spiritual forces such as patience and faith (in the finished work of our Lord and Savior, Jesus Christ, rather than in what man has done such as science, technology and organizations which won't last forever). To avoid sin and error which leads to the death of our body and also our spirit in hell fire, we should let the Word of God be the standard of our lives not AI. If not, God will let us face AI on our own and it will cast the truth down to the ground, it will be the cause of so much destruction like never seen before, it will deceive many and take many captive in order to enslave them into worshipping it and abiding in lawlessness. We can only destroy ourselves but with God all things are possible. God knows us better because He is our Creater and He knows our beginning and our end. Our prove text is taken from the book of John 5:31-44, 2 Thessalonians 2:1-12, Daniel 7-9, Revelation 13-15, Matthew 24-25 and Luke 21. Let us watch and pray... God bless you as you share this message to others.
"clip_embeddings_dim = int(state_dict['image_proj']['proj.weight'].shape[1])
KeyError: 'proj.weight'"
I get this error when adding "Ip adapter Full face" to A1111. Do you know what this error is?
I spent a lot of time researching but still no results!
After deleting and downloading again, I discovered the problem "clip_vision/clip_h.pth" is due to "clip_h.pth" not being updated or simply an error in this file. after replacing the latest version. can work immediately!
@@denle637 Thank you! I had the same problem, and didn't know what to do. Then deleted A1111\extensions\sd-webui-controlnet\annotator\downloads\clip_vision\clip_h.pth , restarted UI, it redownloaded and worked correctly with plugged in ip-adapter-full-face_sd15! ❤❤❤