Just finished watching your latest Comfy UI tutorial, and wow, I'm really picking up a lot from your series. Finally someone who is actually explaining what is behind all the nodes. super helpful! I am an art director and a motion graphic designer who is interested in the new AI field, and would be happy if you can create a tutorial on getting videos to play smoothly with ComfyUI? I mean, making sure each consistent frame is seamless and there's no weird lagging or glitching. That would be a game-changer for a lot of us trying to polish our video projects. Thanks for all the awesome tips so far. Cheers
Is there a way with the controlnet openpose nodes, to load an image of a persons face as it is building the posed model, I can't get that to work. I tried pulid and that was just a mess. I can do it with a prompt, but want to do it with an input image.
Hello, I'd just like to thank you for these videos, they're amazing! Please tell me you're not giving up! I'll introduced your channel to everyone I know who might be interested if you need.
I am brand new to all this, and thank the TH-cam algorithm for showing me your channel. I’ve watched a couple of your videos and subscribed to you, both here and on Patreon. Keep up the fantastic teaching, and hope your channel will reach the 1k subscribers mark very soon. Please share a bit about yourself personally /professionally as well if you don’t mind.
Interesting... What I'm after is the ability to input a default-image of a living-room for example, select the sofa in the livingroom, provide an image of a sofa and have the original sofa replaced with the new one. How would I go about to do this with SD?
I’m planning an interior design video that should more or les cover this. But off the top of my head, I’d use a canny preprocessor to turn the sofa image into line art. Then use a prompt to create a new sofa following the same guide lines, which should give the same shape and form, but change the details of the sofa. Is that what you are trying to do?
@@EndangeredAI I've managed to do this with lineart controlnet and ipqdapter but the issue is that I only want the general angle of the sofa in the ai generated image but keep the form of the sofa I want to replace it with. Right now it gets to similar of the sofa in the ai genrqted image and loose some of its properties.
Thanks for all your tutorials. I've had success with them previously but in this case I'm running up to a problem. I've installed ControlNet, downloaded OpenPoseXL and set up the workflow according to your directions. Unfortunately, once the workflow gets to the KSampler I get a very long error message, too much to share here but the first line is this: ComfyUI_windows_portable\ComfyUI\execution.py", line 153, in recursive_execute Sadly I don't understand what it wants me to do. Perhaps you have an idea where I made a mistake?
@@EndangeredAI yeah i can see the (SDXL Openpose), thought maybe there are more ot controlnet like Sdxl canny to be shared, respecting your busy time, and appreciating the shared knowledge, keep up the good work
Kinda sucks that you didn't offer links on where to find the control net models you use. You show us the official 1.5 models, tell us not to use them but instead use the ones that don't have the default model baked in, then don't say where to actually get those.
I am so sorry it totally slipped my mind that I had not included it. I’m making these videos while travelling for work so I may have forgotten. I’ll get them up asap
ปีที่แล้ว
You will get better results if the image you upload to controlnet is the same size as your SDXL output (1024x1024).
Did I miss something? I beleive everything should have been 1024x1024. The pre processors should have output 1024x1024 canny/openpose
ปีที่แล้ว
@@EndangeredAI I'm talking about the dimensions of the image you uploaded. If the width and height of the image is 1024x1024 px, control net can work more accurately. If you are aware, the proposion gets corrupted while processing the image. I know you didn't explain it well. Otherwise, there is no problem with what you do.
I covered it in a previous video. Let me re watch and find the mistake, everything should have been 1024 input and output, unless I just missed it. Thanks for pointing this out. I had to do a few re shoots and may have made a mistake in the footage. Thanks for pointing it out.
ปีที่แล้ว +1
@@EndangeredAII think you give very good lessons. I wish you continued success. Your efforts are admirable.
Just finished watching your latest Comfy UI tutorial, and wow, I'm really picking up a lot from your series. Finally someone who is actually explaining what is behind all the nodes. super helpful!
I am an art director and a motion graphic designer who is interested in the new AI field, and would be happy if you can create a tutorial on getting videos to play smoothly with ComfyUI? I mean, making sure each consistent frame is seamless and there's no weird lagging or glitching. That would be a game-changer for a lot of us trying to polish our video projects.
Thanks for all the awesome tips so far.
Cheers
Starting video tutorials
Is there a way with the controlnet openpose nodes, to load an image of a persons face as it is building the posed model, I can't get that to work. I tried pulid and that was just a mess. I can do it with a prompt, but want to do it with an input image.
Where can we find the other videos?
Hello, I'd just like to thank you for these videos, they're amazing! Please tell me you're not giving up! I'll introduced your channel to everyone I know who might be interested if you need.
Not at all! I have many more I want to make but the end of the year gets a bit crazy and slows me down! Thank you so much for your interest!
I am brand new to all this, and thank the TH-cam algorithm for showing me your channel.
I’ve watched a couple of your videos and subscribed to you, both here and on Patreon.
Keep up the fantastic teaching, and hope your channel will reach the 1k subscribers mark very soon.
Please share a bit about yourself personally /professionally as well if you don’t mind.
Awesome, thank you! I really appreciate the support! I’m so glad you find the videos helpful!
Great content continue my bro
Thank you!
a link to your workflow would be nice also I'm not finding the openpose pose recognition node?
As I said in another comment, drop by the discord, I´ll Dort you out
Interesting... What I'm after is the ability to input a default-image of a living-room for example, select the sofa in the livingroom, provide an image of a sofa and have the original sofa replaced with the new one.
How would I go about to do this with SD?
I’m planning an interior design video that should more or les cover this. But off the top of my head, I’d use a canny preprocessor to turn the sofa image into line art. Then use a prompt to create a new sofa following the same guide lines, which should give the same shape and form, but change the details of the sofa. Is that what you are trying to do?
@@EndangeredAI I've managed to do this with lineart controlnet and ipqdapter but the issue is that I only want the general angle of the sofa in the ai generated image but keep the form of the sofa I want to replace it with. Right now it gets to similar of the sofa in the ai genrqted image and loose some of its properties.
Thanks for all your tutorials. I've had success with them previously but in this case I'm running up to a problem. I've installed ControlNet, downloaded OpenPoseXL and set up the workflow according to your directions. Unfortunately, once the workflow gets to the KSampler I get a very long error message, too much to share here but the first line is this: ComfyUI_windows_portable\ComfyUI\execution.py", line 153, in recursive_execute
Sadly I don't understand what it wants me to do. Perhaps you have an idea where I made a mistake?
Could you send me a screenshot by email?
That is a very generous offer indeed and I will take you up on it. I will e-mail you later, many thanks. @@EndangeredAI
worth following :)
a ((thank you)) comment for support
looking forward for the links of ControlNet for the comfyUI
Thanks! Are the links not in the description?
@@EndangeredAI yeah i can see the (SDXL Openpose), thought maybe there are more ot controlnet like Sdxl canny to be shared,
respecting your busy time, and appreciating the shared knowledge, keep up the good work
Thank you for covering Controlnet for ComfyUi. Please provide Controlnet models links in your description so that we can find it.
Oh! That’s what I forgot! I knew I was missing something in the description! Thanks for the reminder!
I updated the openpose one in the description. For the other models will need to wait till I get home as I don’t have them handy on my phone
Kinda sucks that you didn't offer links on where to find the control net models you use. You show us the official 1.5 models, tell us not to use them but instead use the ones that don't have the default model baked in, then don't say where to actually get those.
I am so sorry it totally slipped my mind that I had not included it. I’m making these videos while travelling for work so I may have forgotten. I’ll get them up asap
You will get better results if the image you upload to controlnet is the same size as your SDXL output (1024x1024).
Did I miss something? I beleive everything should have been 1024x1024. The pre processors should have output 1024x1024 canny/openpose
@@EndangeredAI I'm talking about the dimensions of the image you uploaded. If the width and height of the image is 1024x1024 px, control net can work more accurately. If you are aware, the proposion gets corrupted while processing the image. I know you didn't explain it well. Otherwise, there is no problem with what you do.
I covered it in a previous video. Let me re watch and find the mistake, everything should have been 1024 input and output, unless I just missed it. Thanks for pointing this out.
I had to do a few re shoots and may have made a mistake in the footage. Thanks for pointing it out.
@@EndangeredAII think you give very good lessons. I wish you continued success. Your efforts are admirable.
@ thank you!
NVM it's name now is openpose pose
not the best tut i have to say