Good that you were able to workout various methods of updatign and fixing the error. I think it makes most sense to learn and use the new IP Adapter V2 and update , change previous workflows
I agree, @HiProfileAI! In the long run, updating and changing previous workflows is the best approach. Although if there are immediate deadlines to meet, pushing the update to a later date is also an option.
I don't get why people have a hard time adapting. Even if I have very complex workflows, it really was not much of an effort to change my workflow nodes. Great work btw. You are a master. I really like the style transfer SDXL.
Because I barely have enough time to use Comfy let alone rework dozens of different workflows and endlessly tweak settings trying to replicate previous results.
It's about how our brain works. I've stayed away from nodes in pretty much every app that uses it. My brain is great at graphics, illustration and any kind of image creation. Nodes drive my brain crazy.
thanks a lot for creating this video!, wracked my head for 4 hours but couldn't solve problem. Your video solved it instantly. Thank you so much. Angel sent form heavens!!
I used the new V2 nodes with my workflows and with a fixed seed and parameters but I can't get the exact picture that I got it before with the old version ....🤔 Then I make a rollback as you explained and I got the exact output from my workflow . So I noticed there are differences between the old and the new nodes where there is no "noise" parameter and some other things. Thanks very much for your awesome and helper tutorials ...... 👍
Thank you so much!, im noob on this and i was really stuck, you really helped me so much and I could complete my workflow withouth problem!, you´ve got another suscriber!
As they say ....a head only a mother would love - however the content is great!! Thanks for continued contributions back to the community....... PS I often find my self waving back to your wave in your intro 😋👋
Hey, just wanted to mention that both inputs for unified loader are optional, it will grab the correct ipa model and clip vision model, if it doesn't you need to rename the models as specified on the GitHub page for the nodes
@@JavierPortillo1Chaining saves VRAM as the loader can reuse/know models are already loaded. The model input is needed for the node to know if you use SDXL or v1.5 and pick the right models.
I waited a little while before updating so I could learn the new stuff. Things are a bit different, but it's not that complicated. The basic stuff is still the same. You don't have to redo your entire workflow either, just replace the affected nodes. The new version is a big improvement, so I would recommend upgrading.
Thanks. Is there also a way to revert back comfyui depencies updates or maybe reset it to the "factory settings"? The last update somehow broke clipseg for me. Using a fresh comfyui fixes this but now of course I need to setup this fresh version again to my liking, which maybe can be avoided if I can revert...
Hello, @DanielPartzsch! You will need to know the version of the dependencies before the updates to do so. You can make a backup copy of the dependencies version by doing "pip freeze > requirements.txt.bak". With this you can revert back. Since you have already updated, I do know of other ways to reset the dependencies.
I'll be upgrading, why hang onto old code when the IPAdapter will be upgraded again in the future to V3 and I have to keep rolling back Comfy to keep my existing workflows working.
Does this rework include a port for AMD graphics card? IP adapter in comfy used to push out errors when I tried it with my AMD 5500 xt, and in the page even face ID said it didn't support AMD, did that change or is it just for nvidia?
Thanks for this tutorial. BTW, could you post or link to your "run_nvidia_gpu.bat"? I use my own, older batch file, I just want to verify that I'm not missing something. Thanks.
You can download the clip_vision models from this link: bit.ly/48VHCzN. Place the models in ComfyUI folder > models > clip_vision folder. After this try again. The link I gave also have instructions on how to download and save other models.
@@CodeCraftersCorner Thanks! Also, don't forget to rename them. They are on the website with a different filename so you have to copy the one from the readme. Not sure why they do this.
There was another able recently where Matteo added more nodes and strong style transfer option. That may have broken your installation. I suggest just upgrading to latest version and rework the workflows. It seems more updates are to come. I understand it is inconvenient but the new features are game changing. If you really need a previous version, you can use the git log --oneline method to find a commit id to rollback to.
@afadfafadfadfadfadfadfa2203! you can try the following: 1. right click on an empty canvas area. 2. Go to Add Node > ipadapter > embeds Try to see which node you need from this list for your workflow.
How they make them out to us and they don't work especially the reactor and this one as well why they don't repair them and they do the more easy and simple For me they waste me a lot of time to do something
@@user-je7qy5ey3y yes, it can be quite tough when learning these latest technology programs - but with anything, patience and willing to learn and understand is important. Many people are willing to help, don't be afraid to ask.
Good that you were able to workout various methods of updatign and fixing the error. I think it makes most sense to learn and use the new IP Adapter V2 and update , change previous workflows
I agree, @HiProfileAI! In the long run, updating and changing previous workflows is the best approach. Although if there are immediate deadlines to meet, pushing the update to a later date is also an option.
Particularly for Tiled features, huge upgrade!
@sorijin Yes, great addition!
I don't get why people have a hard time adapting. Even if I have very complex workflows, it really was not much of an effort to change my workflow nodes. Great work btw. You are a master. I really like the style transfer SDXL.
Because I barely have enough time to use Comfy let alone rework dozens of different workflows and endlessly tweak settings trying to replicate previous results.
It's about how our brain works. I've stayed away from nodes in pretty much every app that uses it. My brain is great at graphics, illustration and any kind of image creation. Nodes drive my brain crazy.
thanks a lot for creating this video!, wracked my head for 4 hours but couldn't solve problem. Your video solved it instantly. Thank you so much. Angel sent form heavens!!
Great to hear!
this update has been very painful, so much workflows broken, your video has been very helpful! thank you very much.
Glad it helped @dasomen!
I used the new V2 nodes with my workflows and with a fixed seed and parameters but I can't get the exact picture that I got it before with the old version ....🤔
Then I make a rollback as you explained and I got the exact output from my workflow .
So I noticed there are differences between the old and the new nodes where there is no "noise" parameter and some other things.
Thanks very much for your awesome and helper tutorials ...... 👍
Thank you for the kind words. Glad the video was helpful, @user-kk2ve1un4u!
Thank you so much!, im noob on this and i was really stuck, you really helped me so much and I could complete my workflow withouth problem!, you´ve got another suscriber!
Glad it helped! Thanks for the sub!
ComfyUI IPAdapter V2 is well optimized. I'll rework all my workflows.
👍
As they say ....a head only a mother would love - however the content is great!! Thanks for continued contributions back to the community....... PS I often find my self waving back to your wave in your intro 😋👋
Thanks for the support, @yiluwididreaming6732! 👋
Hey, just wanted to mention that both inputs for unified loader are optional, it will grab the correct ipa model and clip vision model, if it doesn't you need to rename the models as specified on the GitHub page for the nodes
Also, I think you don't need to chain the loaders, just have one for each IpApadter node, select a preset and leave the inputs empty.
@@JavierPortillo1Chaining saves VRAM as the loader can reuse/know models are already loaded. The model input is needed for the node to know if you use SDXL or v1.5 and pick the right models.
@@sea_ftw Yeah you're right, my mistake.
Thanks for sharing!
Found I had to download the new clip vision models manually, once those were in place all worked. Tense moment when your workflows stop working. 😅
I waited a little while before updating so I could learn the new stuff. Things are a bit different, but it's not that complicated. The basic stuff is still the same. You don't have to redo your entire workflow either, just replace the affected nodes. The new version is a big improvement, so I would recommend upgrading.
Thanks for sharing, @Ulayo! Glad to hear the update went smoothly for you.
it sure is! Once you get it running it really is a powerhouse!
👍
Gracias!!!!! THANKS!!!!!! Saludos desde ARGENTINA
OMG! You literally saved me! Thanks.
Glad to hear it!
thanks a lot for creating this video!
Thx that helps. not do exactly same way btw.
wow you are my hero, thank you so much!
Thanks!
Thank you, @bordignonjunior!
EXCELLENT information!
Glad it was helpful, @Injaznito1!
T H A N K Y O U ! ! ! ! ! ! ! ! ! ! ! !
great video thank you!
Glad it helped!
cant find complete code rewrite even tho i look all of the line.
Hello @kaleabspica8437! You can look at the git diff to see the changes. Here is the link: bit.ly/3TGqKpY
Thanks. Is there also a way to revert back comfyui depencies updates or maybe reset it to the "factory settings"? The last update somehow broke clipseg for me. Using a fresh comfyui fixes this but now of course I need to setup this fresh version again to my liking, which maybe can be avoided if I can revert...
Hello, @DanielPartzsch! You will need to know the version of the dependencies before the updates to do so. You can make a backup copy of the dependencies version by doing "pip freeze > requirements.txt.bak". With this you can revert back. Since you have already updated, I do know of other ways to reset the dependencies.
I'll be upgrading, why hang onto old code when the IPAdapter will be upgraded again in the future to V3 and I have to keep rolling back Comfy to keep my existing workflows working.
👍
nice one, thanks bro
Thank you too, @bookssums!
Why don't I have an IPadapter node? After the update, only IPadapter Ad is available
Hello @onefivethree315! You can try to manually delete the IPadapter folder and re-install in.
Does this rework include a port for AMD graphics card? IP adapter in comfy used to push out errors when I tried it with my AMD 5500 xt, and in the page even face ID said it didn't support AMD, did that change or is it just for nvidia?
Hello, @santiagoarellano8107! I cannot say personally as I do not have an AMD card. You can try posting an issue log on the IPadapter github page.
Thanks for this tutorial. BTW, could you post or link to your "run_nvidia_gpu.bat"? I use my own, older batch file, I just want to verify that I'm not missing something. Thanks.
Thanks @contrarian8870! I use the default from the ComfyUI windows portable installation.
I have upgraded and got better results afterwards so I am not getting back again...
Great!
Have you got the FaceID working properly? No matter where I put the models, the new FaceID node can not find them
reactor node?
Hello @_its.just.regi_! The FaceID from IPadapter is working for me. Check if the model names are defaults one.
hello. great video !!! your videos help a lot. please, mine says clipvision model not found. how do I fix it?
You can download the clip_vision models from this link: bit.ly/48VHCzN.
Place the models in ComfyUI folder > models > clip_vision folder.
After this try again. The link I gave also have instructions on how to download and save other models.
@@CodeCraftersCorner Thanks! Also, don't forget to rename them. They are on the website with a different filename so you have to copy the one from the readme. Not sure why they do this.
Yes, thanks for reminding me.
so its broken again? comfy cant load last version anymore :(
There was another able recently where Matteo added more nodes and strong style transfer option. That may have broken your installation. I suggest just upgrading to latest version and rework the workflows. It seems more updates are to come. I understand it is inconvenient but the new features are game changing. If you really need a previous version, you can use the git log --oneline method to find a commit id to rollback to.
do you know what this is? IPAdapterApplyEncoded, it just says that it is missing
Hello @afadfafadfadfadfadfadfa2203! IPAdapterApplyEncoded is not available in the new version.
@@CodeCraftersCorner oh ok, do you know what it would be replaced with?
@afadfafadfadfadfadfadfa2203! you can try the following:
1. right click on an empty canvas area.
2. Go to Add Node > ipadapter > embeds
Try to see which node you need from this list for your workflow.
cant for the life of me figure out how to work around this ipadapter update, i cant install the old one
do the new models work with automatic1111 ?
Sorry @johnmenezes2031! I have not touch Automatic1111 in quite some time. From what I remember, support for Automatic1111 comes after 1 or 2 weeks.
If you lived by me id five you the gawk gawk 2000 thanks brother.
Thank you, @kingbeard_!
How they make them out to us and they don't work especially the reactor and this one as well why they don't repair them and they do the more easy and simple For me they waste me a lot of time to do something
Yes, it is quite challenging as there are too many things changing.
tutorial install comfyui browser please
Hello @caluxkonde! Yes, I can do this during the weekends.
Request completed: ua-cam.com/video/qKO6pZAv0jM/v-deo.html
@@CodeCraftersCorner thanks lord, ur best
👍
No problem!
For me is a disaster
i have no issues with reactor via insightface/faceid etc now, i used to - i made to sure to update comfyui/reactor/ipadaptor
@@2PeteShakur I'm still learning but I find them certain things very difficult
@@user-je7qy5ey3y yes, it can be quite tough when learning these latest technology programs - but with anything, patience and willing to learn and understand is important. Many people are willing to help, don't be afraid to ask.
Thank you for your encouraging words, @2PeteShakur!
Awesome
thanks