#### links from the Video #### DW Pose: github.com/IDEA-Research/DWPose A1111 1.6 Guide ua-cam.com/video/1Ldc5kZKKFw/v-deo.htmlsi=Hgb0rtVEhLxEQkkL IMPORTANT: For A1111 you do NOT need to download the ZIP file i show in the video. That was only for the pre-release.
It gives error message when I try to install from URL. " Clone succeeded, but checkout failed. You can inspect what was checked out with 'git status' and retry with 'git restore --source=HEAD :/' ' "
Btw if you just found this video, and get an error while installing, know that you don't need to install it anymore, it looks like DWPose actually has been included in A1111 1.6 /w ControlNet options recently, you'll find DW as a pre-processor in the drop-down menu when you select Openpose model.
i keep getting an error everytime i press the exploding button thing oliver talked about, and it wont show a preview of the pose. Doesnt matter if i install the dwpose repository or just use the build in one. It is really wierd
I recommend keeping your ControlNet model files in the sd-webui models/ControlNet directory (not in extensions/sd-controlnet/models). You’ll never have to move them when tinkering with the Controlnet extension and they will be read by A1111 bu default
Which ComfyUI or A1111 folder would you put the two DWPose files in? The GitHub page has a different file structure ( ControlNet-v1-1-nightly/annotator/ckpts); do we simply bung in the two files into extensions/sd-controlnet/models or sd-webui models/ControlNet? Thanks!
thank you, this comments saved me from all the fussy install business. Just choose openpose on controlnet and select dw_openpose_full instead of openpose_full. fpr the pre-processor. Great results!
It gives error message when I try to install from URL. " Clone succeeded, but checkout failed. You can inspect what was checked out with 'git status' and retry with 'git restore --source=HEAD :/' ' "
Dwpose is much better than the default processor. I wonder if someday we will have more accurate open pose skeletons with guides for hips and torso. The current one is way to ambiguous for more complex poses.
Im having an error when installing the DW pose extension on A1111 V1.6 - but I see that the preprocessor is already in there by default? is this why? I get some file and error messages when installing from url.
Hey awesome Video like always! Would love to see a Tutorial about making AI-Videos with like mov2mov or other Methods, where you completely change the Style (from like a Realistic to Anime/Digital Art or smth). And that without a lot of flicker and having clothes and background change all the time! I've seen some people that managed to get it done but I don't really know how?!
Yeah I can get open pose to work, but getting it to open the editor has never worked for me regardless of the SD, A1111, or even python version, it just never works.
Unable to get dwpose preprocessor to work, it just returns error. Is there a specific model that has to be used with it? I only see option to load control-openpose-fp16
I'm confused on the installation. It automatically tries to install DW Pose into ControlNet-v1-1-nightly but when i look up the "nightly" version of control net it says it is not an A1111 extension and not to install it as such.
Big question is does DW Pose address the largest problem with using OpenPose in ControlNEt which is that ControlNet does NOT support depth information in OpenPose. There is no information to determine if a limb is in front or behind another. In OpenPose you can have limbs that are varying widths to indicate perspective but that doesn't work in ControlNet. EDIT: Ok after watching the video more, sadly I see all this new thing is, is a different way of extracting the pose data from an image. It still just generates a FLAT OpenPose image that ControlNet uses and will have the exact same problems. I dont' understand why ControlNet doesn't add support for proper OpenPose skeletons with perspective.
When I click "edit" to open OpenPose editor, I get an error message as an overlay that contains a link to the GitHub for the OpenPose 3D Editor, it doesn't open OpenPose Editor that I already have installed. In fact, if I open OpenPose Editor by itself, it won't allow me to import an image or a background image either. Any suggestions on how to fix this ?
The Openpose Editor doesn't seem to work anymore - when I send a changed pose back to Controlnet from the editor, the pose is shown correctly in the preview window - but then has no effect on the next generating. It sticks to the original pose - so the openpose processing as such works - but not with the changed pose data from the editor! Looks as if some change has been made in the meanwhile which prevents A1111 from respecting the altered pose. Any hints?
Really great, but I'll be the most impressed when they release the versions that are also the rest of the preprocessors (face only, openpose hand, etc). We need the utility of the other preprocessors too, especially openpose hand so it gives us freedom with the face. Also for anyone getting "file name too long" errors, try renaming your root folders to something shorter like just "webui" and whatever is above that. Make note of what you change it too because, tbh, while the comp can update paths when you change a folder name, idk what might not be clear of that and you might find yourself coming up with future unforeseen errors. For me, removing the 255 char limit on filenames in the registry didn't help/work and I still got the same error.
Is there a beginner or an advanced person who can start learning these things? And what about the choice of user interface, 1111 or COMFYUI. What else should a beginner know?
Have you looked into pika labs on discord? It's free and really good. i have some deforum videos on my channel, but i didn't do much with video AI lately
Great tutorial .. but i'm stuck after the installation process. I selected "enable", "Pixel perfect" then checked "OpenPose" and then selected "dw_openpose_full" preprocessor but have no model to select from so the preview remains just the one of the image dont have the "matchstick" version on the right. Surely something i missed ?
@@captbullship Have you figured out where the models are from? I'm assuming they're like checkpoints and need to be downloaded from somewhere before this can be used. I'm trying this as well and only have the option of "None", which is why I imaging it's not working.
is it possible to batch face swap 100 images without uplading 1 by 1 in image to image. I want 1 face in 100 different images without uploading the target image 1 by one and download 1 by 1.
DWPose itself is just a preprocessor, so that is model agnostic. But I don't know if an OpenPose ControlNet model for SDXL is available yet, or if the ControlNet extension for A1111 supports SDXL yet.
I got a really strange problem. When I try to use the DW_pose in A1111 my computer just shuts off, like just goes black. No blue screen or anything. Then it restarts. I can use open pose normally, but as soon as i use the DW_open_pose it shuts completly off. Anyone else with the same problem and/or any solutions?
when I set "dw_openpose_full" and press "run preprosessor" button, the small fire button, my pc restarts. I looked whole internet but could not find anybody that has my issue...
I see the previev image perfectly tracked, but when i render it doesn´t apply the tracked pose to the final image. Where is the problem in this case? I have Controlnet Enabled of course.
@@Felix-qf1ns yeah exactly, but I'm wondering if there's something more architecture specific that could give a still render of a building more realistic life
This is getting real. We're almost to the point of having true digital avatar companions/assistants. Not long ago, this was a topic of science fiction. Soon, you'll be able to use stable diffusion to create the digital avatar and it will become a fully automated AI digital assistant for managing your smart home, driving your car, ordering food, and instantly accessing any information you need whenever you leave the house.
I know it's not directly related to this video, but I installed automatic 1111 1.6 and I can't get Roop to work with it. Has anybody else had a similar issue? Any help would be most appreciated.
So, if you failed to install. you can try: >>>Start Git Bash as Administrator >>> Run command git config --system core.longpaths true it auto switch to onnx branch, but probably failed cause long path. UPD: another part, i've did UPDATE extentions first -> install DW after. (not sure what exactly helped) UPD2: i have to update it again to get 1.1.313 of controlnet.
Back in the day (like 2/3 years ago) I remember AlphaPose was way better then OpenPose, not sure how it compares with current DWPose, but was the best you can get. Wonder if people still use it
Did you use AI to do a final edit on your video @OlivioSarikas? because the video is cropped in weird places. There are times when you are talking and then the video is cropped and you start talking about something else. I'm guessing you are using something to edit out all the pauses or when you say things like "um"? the crops happened in the wrong places.
#### links from the Video ####
DW Pose: github.com/IDEA-Research/DWPose
A1111 1.6 Guide ua-cam.com/video/1Ldc5kZKKFw/v-deo.htmlsi=Hgb0rtVEhLxEQkkL
IMPORTANT: For A1111 you do NOT need to download the ZIP file i show in the video. That was only for the pre-release.
👋
It gives error message when I try to install from URL. " Clone succeeded, but checkout failed. You can inspect what was checked out with 'git status' and retry with 'git restore --source=HEAD :/' ' "
Thank you for great content @OlivioSarikas 🫶🏻
thanks so much, I had postponed updating for a while. Took me a while to figure out the new controls but they soooooo powerful!
Can you transfer a pose to a custom-trained AI model for photorealistic rendering? I want my model to pose in acrobatic ways.
Btw if you just found this video, and get an error while installing, know that you don't need to install it anymore, it looks like DWPose actually has been included in A1111 1.6 /w ControlNet options recently, you'll find DW as a pre-processor in the drop-down menu when you select Openpose model.
i keep getting an error everytime i press the exploding button thing oliver talked about, and it wont show a preview of the pose. Doesnt matter if i install the dwpose repository or just use the build in one. It is really wierd
DW Pose ControlNet 1.4V is not working for now.
@@anderstofthansen7749 I also get that error and cant see the preview
I recommend keeping your ControlNet model files in the sd-webui models/ControlNet directory (not in extensions/sd-controlnet/models). You’ll never have to move them when tinkering with the Controlnet extension and they will be read by A1111 bu default
its quite unbelievable that people aren't doing that by default :/
For those interested, the DWPose preprocessor also works great for ComfyUI and controlnet.
Which ComfyUI or A1111 folder would you put the two DWPose files in? The GitHub page has a different file structure ( ControlNet-v1-1-nightly/annotator/ckpts); do we simply bung in the two files into extensions/sd-controlnet/models or sd-webui models/ControlNet? Thanks!
how to configure it in comfyUI or SwarmUI (what files to download and how to configure workflow)?
dw pose seems to be preinstalled with A1111 v.1.6.0
thank you, this comments saved me from all the fussy install business. Just choose openpose on controlnet and select dw_openpose_full instead of openpose_full. fpr the pre-processor. Great results!
It gives error message when I try to install from URL. " Clone succeeded, but checkout failed. You can inspect what was checked out with 'git status' and retry with 'git restore --source=HEAD :/' ' "
Another great one Olivio! ControlNet has changed my entire perspective of what I can do w/SD. Really appreciate videos on this powerful tech.
Will it allow us use of the SDXL models in ControlNet yet?
This version of open pose is AMAZING. You are an amazing teacher. Thank you.
Thanks for the video. Hope ControlNet will be soon available for SDXL
Using this with something like Character Creator to make pose images makes things so much easier.
Dwpose is much better than the default processor. I wonder if someday we will have more accurate open pose skeletons with guides for hips and torso. The current one is way to ambiguous for more complex poses.
you can always add depth or something like that to controlnet and add more information about the pose.
Much better than openpose preprocessor. Thank you Olivio for sharing.
DW pose is now already preinstalled if you update your ControlNet :)
Thank you for the detailed and clear step-by-step instructions!
I've been waiting for this for a very long time!!
Im having an error when installing the DW pose extension on A1111 V1.6 - but I see that the preprocessor is already in there by default? is this why? I get some file and error messages when installing from url.
Did you solve that issue? If yes please tel me how because i am having the same problem
DW Pose ControlNet 1.4V is not working for now.
Wow i really like that you can alter the points. Great video thanks.
Still no ControlNet for SDXL. So sad :(
This is exactly what I need, gonna try it out later
Thanks for this nice Extension .. i would love to see more helpful extension videos in the future 🔥👍🏻
Very nice. Thanks for highlighting it Olivio.
Is this able to work with the new version and SDXL? I thought ControlNetSDXL was still not availible yet?
This doesnt work for me with sdxl
Hey awesome Video like always! Would love to see a Tutorial about making AI-Videos with like mov2mov or other Methods, where you completely change the Style (from like a Realistic to Anime/Digital Art or smth). And that without a lot of flicker and having clothes and background change all the time! I've seen some people that managed to get it done but I don't really know how?!
Enigmatic_E or Tokyojab
Thanks for the tutorial, DW works great for me!
Not working, getting instaling error
When I click the Openpose edit window, its just a black window. Any ideas?
Open pose never worked properly make me disappointed, this is great thing to replace it. Thanks for sharing
Yeah I can get open pose to work, but getting it to open the editor has never worked for me regardless of the SD, A1111, or even python version, it just never works.
Available for ComfyUI ? SDXL ?
Edit : Available in ComfyUI, no idea about SDXL
I'm sorry, can you answer. During your conversation at 4:00 minutes. You recommend renaming the folder. What name do I need for the new folder?
Any? newfolder, 123, aijuwhe
Oh man... This is amazing but it's making my processing time 5X longer.
Unable to get dwpose preprocessor to work, it just returns error. Is there a specific model that has to be used with it? I only see option to load control-openpose-fp16
I have a question, my DWPose preprocessor doesn't work error comes up everytime i click the preprocessor button
I'm confused on the installation. It automatically tries to install DW Pose into ControlNet-v1-1-nightly but when i look up the "nightly" version of control net it says it is not an A1111 extension and not to install it as such.
Big question is does DW Pose address the largest problem with using OpenPose in ControlNEt which is that ControlNet does NOT support depth information in OpenPose. There is no information to determine if a limb is in front or behind another. In OpenPose you can have limbs that are varying widths to indicate perspective but that doesn't work in ControlNet.
EDIT: Ok after watching the video more, sadly I see all this new thing is, is a different way of extracting the pose data from an image. It still just generates a FLAT OpenPose image that ControlNet uses and will have the exact same problems. I dont' understand why ControlNet doesn't add support for proper OpenPose skeletons with perspective.
How do you apply open pose to an existing image? (generated in Mid Journey) I cant seem to find how to do this
When I click "edit" to open OpenPose editor, I get an error message as an overlay that contains a link to the GitHub for the OpenPose 3D Editor, it doesn't open OpenPose Editor that I already have installed.
In fact, if I open OpenPose Editor by itself, it won't allow me to import an image or a background image either.
Any suggestions on how to fix this ?
I have the same problem...(
I click edit and nothing happens, I can open it with a second link and pose it but can't upload the .json file to control net.
It's a great extension, but it doesn't seem to be able to recognize more than one person in the picture.😢
The Openpose Editor doesn't seem to work anymore - when I send a changed pose back to Controlnet from the editor, the pose is shown correctly in the preview window - but then has no effect on the next generating. It sticks to the original pose - so the openpose processing as such works - but not with the changed pose data from the editor! Looks as if some change has been made in the meanwhile which prevents A1111 from respecting the altered pose. Any hints?
Really great, but I'll be the most impressed when they release the versions that are also the rest of the preprocessors (face only, openpose hand, etc). We need the utility of the other preprocessors too, especially openpose hand so it gives us freedom with the face.
Also for anyone getting "file name too long" errors, try renaming your root folders to something shorter like just "webui" and whatever is above that. Make note of what you change it too because, tbh, while the comp can update paths when you change a folder name, idk what might not be clear of that and you might find yourself coming up with future unforeseen errors.
For me, removing the 255 char limit on filenames in the registry didn't help/work and I still got the same error.
git config --global core.longpaths true
how is it possible to train my own pre-processor like this, for 2d character monsters that are not human shape?
Where can I get that Edit button next to the open pose skeleton? I have the plugin but i don't see the edit button
6:40
Im going nuts over the refiner and how some models are better for poses then letting another take over for detail is just *chef's kiss* and now this?
Please explain, which models are better for poses? Which models take over for details?
Did like he in the video but there a no models i can select from, only the preprocessors. Any idea how to fix ?
Tested with some stubborn images I have. It definitely work better
ok i dont get it. i did the same as u did here. but everytime SD generate a picture with a different pose as the one i use in controlnet
Thanks for the tutorial!
Is there a beginner or an advanced person who can start learning these things? And what about the choice of user interface, 1111 or COMFYUI. What else should a beginner know?
Hey awesome Video like always! Would love to see a Tutorial about making AI-Videos 👍👍👍
Have you looked into pika labs on discord? It's free and really good. i have some deforum videos on my channel, but i didn't do much with video AI lately
I llke the way you pronounced yaml files.
No matter what I do, I follow the steps I see on youtube videos & SD only pops up with TypeError: 'NoneType' object is not iterable
Olivio hi! this dw_openpose works only in sd 1.5 or sdxl too?
Great tutorial .. but i'm stuck after the installation process.
I selected "enable", "Pixel perfect" then checked "OpenPose" and then selected "dw_openpose_full" preprocessor but have no model to select from so the preview remains just the one of the image dont have the "matchstick" version on the right.
Surely something i missed ?
**UPDATE** I got the "matchstick" editor to work(had to restart the whole engine) .. but where can I get "poses" for the model ?
@@captbullship Have you figured out where the models are from? I'm assuming they're like checkpoints and need to be downloaded from somewhere before this can be used. I'm trying this as well and only have the option of "None", which is why I imaging it's not working.
Ugh, I keep trying to install these with the links, but it keeps not showing that it's installed, even after updating and restarting the UI.
is it possible to batch face swap 100 images without uplading 1 by 1 in image to image. I want 1 face in 100 different images without uploading the target image 1 by one and download 1 by 1.
Is it possible to run this in ComfyUI?
Nice thank you again! I assume this is not for sdxl and only for 1.5? or is it?
DWPose itself is just a preprocessor, so that is model agnostic.
But I don't know if an OpenPose ControlNet model for SDXL is available yet, or if the ControlNet extension for A1111 supports SDXL yet.
@@phizc right, it's only a pre processor. Thanks! And no, controlnet for 1111 sdxl is not available yet, so you are right again
how to use it with SDXL?
I got a really strange problem. When I try to use the DW_pose in A1111 my computer just shuts off, like just goes black. No blue screen or anything. Then it restarts. I can use open pose normally, but as soon as i use the DW_open_pose it shuts completly off. Anyone else with the same problem and/or any solutions?
Thank you very much, This is super helpful
when I set "dw_openpose_full" and press "run preprosessor" button, the small fire button, my pc restarts. I looked whole internet but could not find anybody that has my issue...
strange. never heard that. maybe delete controlent and reinstall it (make sure you backup the controlnet models and don't delete them by accident)
Did you manage to solve it? I'm having this problem too.
Does this work with SDXL? I tried a around a bunch and couldnt get it work with various checkpoints, in SD1.5 it works great tho.
Hi there! Can you get this working with comfyui?
I see the previev image perfectly tracked, but when i render it doesn´t apply the tracked pose to the final image. Where is the problem in this case? I have Controlnet Enabled of course.
Thank you very much!
Tell me, please, is changing the pose in open space possible only in txt2img? is it possible in img2img in any way?
I have this error:
ModuleNotFoundError: No module named 'mmengine'
can you help?
is there a way to save/export only the preprocessor results?
Can u make tutorial how to make good quality animations?
Hi Olivio, is there a tool to bring architecture still renders to life with a small animation? Thank you 😊❤ great content
You mean like animatediff?
@@Felix-qf1ns yeah exactly, but I'm wondering if there's something more architecture specific that could give a still render of a building more realistic life
@@Digital_Architectsthere’s a community that’s starting to make modified motion modules for animate diff, it needs more attention.
git config --global core.longpaths true
Can you install for ComfyUI?
I can get DWpose working however I dont see a controlnet for in Comfy
How would the steps change if using comfyUI
This is getting real. We're almost to the point of having true digital avatar companions/assistants. Not long ago, this was a topic of science fiction. Soon, you'll be able to use stable diffusion to create the digital avatar and it will become a fully automated AI digital assistant for managing your smart home, driving your car, ordering food, and instantly accessing any information you need whenever you leave the house.
I know it's not directly related to this video, but I installed automatic 1111 1.6 and I can't get Roop to work with it. Has anybody else had a similar issue? Any help would be most appreciated.
Same, doesn't work. Using ReActor.
Hey, did you resolved the problem? I find out that changing set SKIP_VENV=0 to in environment.bat helps. Hopefully it'll help u too. 👍
@@maraderchikXD thanks, I will give it a go. Still using 1.5 at the moment
my restore faces has mysteriously disappeared, does anyone know why?
So, if you failed to install. you can try:
>>>Start Git Bash as Administrator
>>> Run command git config --system core.longpaths true
it auto switch to onnx branch, but probably failed cause long path.
UPD: another part, i've did UPDATE extentions first -> install DW after. (not sure what exactly helped)
UPD2: i have to update it again to get 1.1.313 of controlnet.
My PC crashes and restarts (blue screen) when I use DW Pose, normal open pose works correctly without crashes. Any solutions?
What kind of hardware are you using?
is there any way to use in confy ui
Can you use DW Pose also for the full body or just for "selfies"?
Hi Sir, one question.
should I check controlNet in img2img ? if i want to upsacle my pict
Only if you haven't already generated an image. If you only want to upscale an image you don't need controlnet (maybe for tile but not openpose)
I'm curious about your setup. Sometimes it looks like you are using Mac OS. Can you enlighten me?
Excellent as usual. Thanks !
Is there a 3D rig with the face somewhere ?
when installing from url, i got the file name too long error huh??
Is this not available for comfyui?
tried that it seems to completely ignoring the pose
Are you using SDXL? it only works with SD1.5
Cool, good video!
the link doesn't appear to work when I try to install it
hello ! can i use it on Invoke AI ?
Thanks for the info :D
Back in the day (like 2/3 years ago) I remember AlphaPose was way better then OpenPose, not sure how it compares with current DWPose, but was the best you can get. Wonder if people still use it
Not sure why I cant save your videos to a list. :(
Did you use AI to do a final edit on your video @OlivioSarikas? because the video is cropped in weird places. There are times when you are talking and then the video is cropped and you start talking about something else.
I'm guessing you are using something to edit out all the pauses or when you say things like "um"? the crops happened in the wrong places.
actually ignore that, I just played parts back and they seem OK now ... I guess I just had a weird caching issue my side
thank you ❤
openpose not working sdxl a1111 1.6v
영상 감사합니다~!
Ummmm.... wonder if is better than tile for anime videos creation...
Update: The answer is no.
Thanks!
ua-cam.com/video/30b2k1p2CiE/v-deo.html "Nicely" she has legs of different length... that is normal?
yay, my SD Benchmark says: 0.04 it/s
yes trained images