ComfyUI 16 Controlnet + IPAdapter Workflow (free download)
Vložit
- čas přidán 20. 01. 2024
- Controlnet ( • ComfyUI 12 Controlnet ... ) and IPAdapter ( • ComfyUI 15 Styling wit... ) can be combined in one ComfyUI workflow, which makes it possible to steer both the composition and the style & atmosphere of the AI generated images. Together with latent- and image upscaling and with face restore it makes for a great default workflow.
Link to the workflows from the video:
drive.google.com/drive/folder...
If there's trouble with the Latent Upscale node, try this one:
github.com/city96/SD-Latent-U...
Link to a large collection of PNG image workflows:
drive.google.com/drive/folder...
Link to video on ComfyUI Manager:
• ComfyUI 03 - ComfyUI M...
Large collection of upscalers:
openmodeldb.info/ - Jak na to + styl
One of the best detailed ai explaining channel. Thank you so much for your hard work, knowledge and free workflows.
Thanks for your kind words, it's appreciated.
Excellent video. Like your style. Thanks for the education and the workflows, they're very helpful! Happy to find you!
This was so informative… i got exactly what i needed which is quite rare on youtube ❤❤❤… thank you and ur voice was so calm and fun… loved every bit of this video
Thank you for your nice feedback, it's appreciated.
I like all of your tutorial videos. Thnkas a lot for always sharing the workflows.
Glad you like them, thx for your feedback, it's appreciated.
Thanks a ton for your contribution to spreading information about how to use this kinda stuff. REALLY appreciate it!
You're welcome, thanks for your feedback, it's appreciated.
You have such a brilliant way of explaining things. This has to be the best content on ComfyUI that I've seen so far. Your workflow is so well ordered and methodical and the results of your experiments are truly an eye-opener. I wonder if you have used the ComfyRoll Multi-ControlNet Stack to mix multiple ControlNets and whether it is preferable to adjust the image with one ControlNet at a time, rather than mixing weights from different ControlNets?
Thanks for your kind feedback, it's appreciated. Yes, I have tried the multi ControlNet Stack, but I found that I seldom have the need for more than one Controlnet at a time, I seldom use it. I can't tell if it is preferable to stack or not to stack ... I bet a couple of experiments can tell us.
@@rudyshobbychannel I feel the same way. For the most part, I think one ControlNet is sufficient to get the result, especially if lines are all you really need to guide the picture. I am just wondering what are the possible use cases for using a ControlNet Stack. Thank you again for making your tutorials. I really enjoy your straightforward style, and of course, the clear explanation of the concepts.
@@brgtubedev001 There are Controlnets that are about color, that can be a possible stack, but I rather do that with the IPAdapter. And I bet one could use a Canny for a room- or city layout and add an Openpose to put a character in.
@@rudyshobbychannel Yeah, I just came across the recolor ControlNet, that's an interesting one to try out. And I can sort of see where you are going with mixing a Canny with Openpose. That would be an interesting workflow indeed. Would it be right to say that instead of overlaying a character in, you would use this method to instead generate the character right in the scene?
@@brgtubedev001 I'll give it a try one of these days.
Wonderful, thanks!!!!
You're welcome. Thanks for the feedback.
Lekker bezig Rudy (JDx)
Thx
I have watched many comfyui videos on youtube, this is definitely the best comfyui tutorial, thanks for your generous sharing. I encountered an error while loading the json file, it says textbox node is missing. But when I check in the comfyui manager, there is no textbox node. The error message actually points to the "Filename" box which is red. May I ask how to resolve this issue? thanks.
That node is part of the Derfuu mods: github.com/Derfuu/Derfuu_ComfyUI_ModdedNodes. It is meant to enter a filename for the images that can be saved in several nodes. You can also delete that filename node and then convert the inputs it was connected to, to widgets.
@@rudyshobbychannel thanks
Can you help me fix the Filename node? It is red, but it doesn't show up as missing under the manager. Much thanks!
It is part of this node collection: github.com/rgthree/rgthree-comfy Any other text/string output node will do by the way, and you could also just change the inputs it is connected to into widgets and then delete the node.
Hi, thank you for sharing your workflows, which json file should I use?
Controlnet IPAdaper (Turbo)
@@rudyshobbychannel Thanks! after installing missing nodes it tells me I'm still missing LatentUpscaler and I can't find anywhere to download it.
@@nekola203 I think the one I used is part of the so called Efficiency Nodes. But you can also use any other Latent Upscale node that you can find and insert it at that spot.
@@rudyshobbychannel great thanks again!
I think this one should do the job too:
github.com/city96/SD-Latent-Upscaler
[IPAdapter] The updated 'ComfyUI IPAdapter Plus' custom node has undergone significant structural changes, rendering it incompatible with existing workflows.
I loaded your workflow and I get an error "InsightFace must be provided for FaceID models" anyone know what might be wrong??
It could be that you do not yet have all models or other resources needed. Carefully follow the guidelines on the IPAdapter Github site: github.com/cubiq/ComfyUI_IPAdapter_plus
unable to find out text box node
It is part of the Derfuu Nodes, can be installed via the Manager or via Github. You can also convert the inputs to widgets in the three nodes it is connected to and then delete it.
I generated woman face that i very like. Is any way to apply this face to rest of images which i will generate? Reactor or other face replace nodes makes face very bluury when image is bigger than 512x512.
IPAdapter could help, but I doubt if it can do a better job than Reactor. New tools are on the way, like this:
github.com/ZHO-ZHO-ZHO/ComfyUI-PhotoMaker
czcams.com/video/CYx6I0OetF4/video.htmlsi=98WjacGC-h5NAh0K
@@rudyshobbychannel I tried to use this PhotoMaker but i got very long error starts with:
Error occurred when executing PhotoMakerLoader:
Error(s) in loading state_dict for PhotoMakerIDEncoder:
I tried IPAdapter but face is blurry or not much similar to source :(