This is cool and all, but I feel like its beating around the bush of what everyone ultimately wants - a fully posable character model within stable diffusion, hands, feet, and skeleton. GIven how fast this stuff is coming out, I can only hope something like this comes soon!
Great as always! 👍 Also after a lot of this is fun, this is huge, this is insane you went back to a rather informative thumbnail which is very much appreciated 🙏
Waiting for your next video on the update/alternate version of the Latent Couple extension that allows you to directly paint masks by color to determine which sub-prompt goes where .. basically kinda like Nvidia Canvas! I tried it and it worked, now I have two versions of the extension in the same drop menu (in case you want to use the original version with the divisions for some reason).
I downloaded the latest commit of sd-webui-controlnet and the controlnet tabs (Control Model - 0, Control Model - 1, ..) were missing. You have to enable it in settings.
@@Aitrepreneur By the way - have you come across Multi ControlNet messing up your images' faces? Because it seems to do that with mine. I've played with it a couple of days. And I can't seem to figure out what's doing it. But the quality of my images and details in the images look terrible when I apply more than one controlnet. Did that happen to you too? It's both locally on my own laptop - but also on sites like RunDiffusion and RunPod...
this one is good, because you dont need long time to pose the hand manually, it really helpfull and the 3d map fix the proble that open pose have that it dont have 3d prepective
Looking forward to when you can map the software on top of the other open source software that allows you to track hands and movement from normal cameras
thank you for the videos! I think it's only a matter of time until a new extension for manipulating the fingers come along for Stable Diffusion too :D :D
I can't get posex to work at all. I get the posex tab, the only thing that shows up is the checkbox that says "Send this image to ControlNet". Am I missing a prerequisite extension or something?
Just an an TLD/ FYI for those who can't be arsed with all of this Aitrepreneur covered a faster way (search for EASY POSING FOR CONTROLNET). One thing that puzzles me is what the head stalks represent in the skeleton, eyebrows ?
Jesus goddamn christ. I turn around for 1 damn second and there's already a new extension that gets automatic 1111 closer to clipping my toenails and wiping my ass for me😀👍👍👍.
Sadly for me all work but when i generate the images the fingers stills deformed, the pose and the hands posture are perfect but the fingers still looking bad for me. I dont know why i have the same settings as you... Maybe is because im using stable diffusion 1.4?
Just installed posex and it doesn't show an image of the skeleton, only one line of text. Does it not work with the latest version? Update: Use 3D openpose instead it works wonders
hi there thank for the video its amazing , I am doing the same steps you explained in you video but somehow the depth extension disappeared and wont show a gain ant advice please
Hi, how can I make appear the Tabs Control Net - 0 - 1 - 2. In my end dont have those and have it updated with all the requeriments, and gradio updated too
nvm I solved it, had to setup the parameters in the SETUP of the control net and set multi control net, and add 2 o 3 more :D I let it here if someone else face this issue.
When I open and close the PoseX window (a few times), the interface freezes and no button responds to clicks. I have to restart the web UI to unfreeze it. Any solution?
@@Aitrepreneur Hi, I have tried on 2 different PCs to install it, there no 3d box with a squeletton appearing, checking the "send to controlnet" does not change anything. seems many people have this issue!
neat, only problem i see is that the 3d space is controlled only by the camera, so you can’t have different 3d perspective for two different open pose skeletons. sure they will have different poses but they will always have the same perspective which is kind of useless for more than one character. instead of of a camera, or in addition to the camera, a rotate skeleton (in 3d space) would be useful
What's next? FEET? Hey curious, are those textual inversions subject to the same pickle stuff? I think those pickle scanners only scan the checkpoint model folders right?
Thanks for the tutorial, it is amazing!! Do anyone know why I tried to send the image to control net from depth library, it just sent the hand but not the open pose?🤔🤔
Unfortunately I can not seem to get the pose x to run properly inside of stable diffusion, I tried deleting and reinstalling it as well as running the python file inside of it. Running the python file seemed down download what it mentioned in the requirements file but gave me a separate local address to run it from, which I guess is a work around since it will not properly show up but I seem to be able to use it.
Hi AI i cant seem to install the models into controlnet, i only get the openpose thing on the preprocess side but it doesnt work there. any ideas on what to do? :(
I feel like i will need a New GPU in the Not so far future :') Do you have any recommendations for good gpus in the low to middle price section? (Max 450 €)
Your videos are amazing but I’m having trouble. I updated my web ui and everything broke. I know you made a video with info on how to revert back to a previous version using the hash number but I’ve been looking for two hours and I cannot find it. Please help!!
K! Are you in contact with any of these wizards creating this stuff? I'm still desperate for a tool that allows you to tell it where a character is looking... including eyes. It so desperately wants the characters to look at the camera 95% of the time.
@@eldrickzero4885 Yeah. That's the easy one. Have you tried to get the character looking anywhere else? Every once in a while maybe. But never at the same thing.
super cool, but blender is better for posing with ik, and this plugin doesn't keyoff the nose and eyes, which is required to make character face away from camera
I can't seem to rotate hands in the depth map library. Anyone got any clue what to do? The top ultimate top square only displays a "stop sign" icon when hovered.
Guys Deph Library tab dont appears in my UI and also the Posex tab!! its a BUG? i restart the UI and my PC but its still not available, the extensions are installed but are not visible.... Someone know what happen??
Is it possible to create anime style characters with this? I tried using the blender 3d model for quite a few Anime models and it always tried to make the face Look realistic… is it the same for this method aswell or is it also trained for anime?
He covered that in on of his previous videos. I think it's the one that says "This Is Huge" in the thumbnail and has two girls with their hands in front of their face
Everyone is complaining in the recent comments and yet you ignore them. What about warning us in the title that the extension is outdated and broken instead of ruining people's stable diffusion? Dislike.
HELLO HUMANS! Thank you for watching & do NOT forget to LIKE and SUBSCRIBE For More Ai Updates. Thx
I'm having trouble creating my own depth maps by hand. It would be interesting for you to create a tutorial.
This is exactly what was talking about the other day! Of course you have more freedom with Blender but this is the huge time saver! Thank you K!
the phase at this point of things we actually wanted with A.I is stunnig, cant wait for new stuff😮😮😮
This is cool and all, but I feel like its beating around the bush of what everyone ultimately wants - a fully posable character model within stable diffusion, hands, feet, and skeleton. GIven how fast this stuff is coming out, I can only hope something like this comes soon!
:)
Yeah, the hands thing is basically just a clipart. But this is progress
All of these new extensions are the process it's taking to work towards a full 3D model. Good tech takes time.
I give it 2 months.
Sure, but that's just how these things go. Eventually we'll get it built in, but this is a pretty simple (if somewhat tedious) solution.
Great as always! 👍
Also after a lot of this is fun, this is huge, this is insane you went back to a rather informative thumbnail which is very much appreciated 🙏
omg, deserve a comment and like. and on top of that super simple explanation.
Waiting for your next video on the update/alternate version of the Latent Couple extension that allows you to directly paint masks by color to determine which sub-prompt goes where .. basically kinda like Nvidia Canvas!
I tried it and it worked, now I have two versions of the extension in the same drop menu (in case you want to use the original version with the divisions for some reason).
Yes I'm working on it :)
Hmmm what ? Name of that extension please~~?
@@tetsuooshima832 It's the (ashen-sensored/stable-diffusion-webui-two-shot)
Omg….
@@AscendantStoic thanks, I found a tutorial by Sebastian Kamph to install it properly, it works !! Awesome !
Your videos are super helpful and my favorite/best resources for all things Ai! Thanks, keep it up!
Thanks I appreciate it! ;)
I downloaded the latest commit of sd-webui-controlnet and the controlnet tabs (Control Model - 0, Control Model - 1, ..) were missing. You have to enable it in settings.
I was looking at the Posex documentation just a few minutes ago trying to figure out how it's different than the OpenPose Editor. Thanks!
Enjoy ;)
Yes yes, it was a matter of time! Thank you!
You bet!
You are absolutely awesome. Thank you. I've been waiting for this. So awesome!🙏
Happy to help!
@@Aitrepreneur By the way - have you come across Multi ControlNet messing up your images' faces? Because it seems to do that with mine. I've played with it a couple of days. And I can't seem to figure out what's doing it. But the quality of my images and details in the images look terrible when I apply more than one controlnet. Did that happen to you too? It's both locally on my own laptop - but also on sites like RunDiffusion and RunPod...
I had empty posex ui only "send image to controlnet"
Very, very super cool! ....WOW!... Big FANX!!!!
The speed of changes going on now is kind of insane
this one is good, because you dont need long time to pose the hand manually, it really helpfull and the 3d map fix the proble that open pose have that it dont have 3d prepective
Exactly what I needed!
the latest Pose version does not work with the latest WebUI, no fix so far, no 3D Panel visible, one needs to rollback to the prev versions
Im having a lot of issues as well with everything updated.
Game changer
I have to agree :)
The game changes so often we might as well be playing Calvinball
Hahaha game changer, we do that from 30 years
@@albertoUnreal As someone who's 32 years old, I don't remember AI text to image generation existing when I was a kid 🤔
@@IceMetalPunk AI exists from 22 years true
THIS PROGAM IS MOVING SO FAST.
It's getting gooder and gooder...This is exciting stuff!
Looking forward to when you can map the software on top of the other open source software that allows you to track hands and movement from normal cameras
In what way do you mean? You can already generate both an OpenPose skeleton and a depth map from a regular camera image.
thank you for the videos!
I think it's only a matter of time until a new extension for manipulating the fingers come along for Stable Diffusion too :D :D
If it was 3d model preview window and transformer, it would be nice for sure :)
For now daz3d is winner
Do you know How mich ram is needed for this?
The latent couple mask extension for. Example didnt Worked for my 6gb RAM...
I can't get posex to work at all. I get the posex tab, the only thing that shows up is the checkbox that says "Send this image to ControlNet". Am I missing a prerequisite extension or something?
Click that checkbox and it should appear
Now we just need a perspective view, locked limb-sizes and working IK and auto-balancing :P
I barely get done testing one extension and there's already a new, improved tool available.
Just an an TLD/ FYI for those who can't be arsed with all of this Aitrepreneur covered a faster way (search for EASY POSING FOR CONTROLNET). One thing that puzzles me is what the head stalks represent in the skeleton, eyebrows ?
Awesome, next evolution will be holding object depth map library right ?
I have the extension installed but the poser does not display when I click the drop down.
Me too
And me
Click the send image to control net, then it will appear
Could we use the same thing for feet also? 😜
Great video, but one question-- I see how you can grab the points and move them, but how do you rotate them, like making the head or waist rotate?
Jesus goddamn christ. I turn around for 1 damn second and there's already a new extension that gets automatic 1111 closer to clipping my toenails and wiping my ass for me😀👍👍👍.
Hahahaha :D
all hail our lord and savior Ai Overlord!
Sadly for me all work but when i generate the images the fingers stills deformed, the pose and the hands posture are perfect but the fingers still looking bad for me. I dont know why i have the same settings as you... Maybe is because im using stable diffusion 1.4?
Are there any depth map libraries of hands?
My man really using Abyss orange Mix Hard for his demos 😂
I have no idea what this is tbh I was just recommended this model for some reason
@@Aitrepreneur Surely 😂
I mean... isn't is clear with 2.0 that *"that stuff"* makes normal image generation better?
Just installed posex and it doesn't show an image of the skeleton, only one line of text. Does it not work with the latest version?
Update: Use 3D openpose instead it works wonders
Thanks
can you make a video showing the best way to take a pre existing image with bad hands and fix them using these tools??
hi there thank for the video its amazing , I am doing the same steps you explained in you video but somehow the depth extension disappeared and wont show a gain ant advice please
why the heck dont you number them?? Or write update in title?
You are the best
Thanks! I Appreciate it ;)
Hi, how can I make appear the Tabs Control Net - 0 - 1 - 2. In my end dont have those and have it updated with all the requeriments, and gradio updated too
nvm I solved it, had to setup the parameters in the SETUP of the control net and set multi control net, and add 2 o 3 more :D I let it here if someone else face this issue.
When I open and close the PoseX window (a few times), the interface freezes and no button responds to clicks. I have to restart the web UI to unfreeze it. Any solution?
MVP
Posex doesn't work for me for some reason. Installed, applied, I can see the tab, but it's empty.
Click on the checkbox "send to controlnet" it should appear
@@Aitrepreneur Thank you!! Also had this issue
@@Aitrepreneur Hi, I have tried on 2 different PCs to install it, there no 3d box with a squeletton appearing, checking the "send to controlnet" does not change anything. seems many people have this issue!
neat, only problem i see is that the 3d space is controlled only by the camera, so you can’t have different 3d perspective for two different open pose skeletons. sure they will have different poses but they will always have the same perspective which is kind of useless for more than one character. instead of of a camera, or in addition to the camera, a rotate skeleton (in 3d space) would be useful
Give it a week 😅
@@AZTECMAN Day* at this rate
3D mode doesn't work
What's next? FEET?
Hey curious, are those textual inversions subject to the same pickle stuff? I think those pickle scanners only scan the checkpoint model folders right?
Probably nothing prevents you from using depth images of feet in that hand library, of course would be better to have categories for each body part.
help, when I pass it to controlnet the doll that he designs does not appear, that is, it only appears with a black background. any solution?
Thanks for the tutorial, it is amazing!!
Do anyone know why I tried to send the image to control net from depth library, it just sent the hand but not the open pose?🤔🤔
hi, thanks, but I don't get a squeletton image posex, there is just "send this image to controlnet" and target control number! very disapointing..😁
same here... did you manage to fix it ?
Good stuff, they should add feet too...
Unfortunately I can not seem to get the pose x to run properly inside of stable diffusion, I tried deleting and reinstalling it as well as running the python file inside of it. Running the python file seemed down download what it mentioned in the requirements file but gave me a separate local address to run it from, which I guess is a work around since it will not properly show up but I seem to be able to use it.
How can I have different tab for different Control Model (like Model - 1, Model -2, etc.)?
How I solve this error? "AssertionError: extension access disabled because of command line flags" I can't install any extension
Hi I already follow the tutorials but on My posex panel didn't show any skeleton that I can edit. Anybody know the problem?
Hi AI i cant seem to install the models into controlnet, i only get the openpose thing on the preprocess side but it doesnt work there. any ideas on what to do? :(
omg 😂 this is so easy
any idea of how I can train a model to use a 3d render and then add it to stable diffusion to ask for things with the 3d model?
💞🙏💞 Thank you so SO much.
Any time!
I feel like i will need a New GPU in the Not so far future :')
Do you have any recommendations for good gpus in the low to middle price section?
(Max 450 €)
Your videos are amazing but I’m having trouble. I updated my web ui and everything broke. I know you made a video with info on how to revert back to a previous version using the hash number but I’ve been looking for two hours and I cannot find it. Please help!!
git reset --hard [hash]
Now we just need the Save PNG to save to the Stable Diffusion Outputs Folder as a new Folder "Open Pose."
when I open the poseX tab, only the box to send the image to controlNet appears. Just it
Im install the extensions, but i cant see it on the WEB UI!! what happen?¿?¿
K! Are you in contact with any of these wizards creating this stuff? I'm still desperate for a tool that allows you to tell it where a character is looking... including eyes. It so desperately wants the characters to look at the camera 95% of the time.
thats a tag looking_at_viewer
@@eldrickzero4885 Yeah. That's the easy one. Have you tried to get the character looking anywhere else? Every once in a while maybe. But never at the same thing.
super cool, but blender is better for posing with ik, and this plugin doesn't keyoff the nose and eyes, which is required to make character face away from camera
what is ik?
I wonder how well this works with composable LoRA, if at all
I can't seem to rotate hands in the depth map library. Anyone got any clue what to do? The top ultimate top square only displays a "stop sign" icon when hovered.
How do you get multiple Control Models in ControlNet? Mine only doesn't show "Control Model-0 Control Model-1 Control Model-2"
Settings, controlnet, enable multicoltrolnet as you desire, apply and restart the ui
@@Nyxo1000 thank you!!
how can i add multiple tabs to my control net webui to add the pose and the hand depth together
Guys Deph Library tab dont appears in my UI and also the Posex tab!! its a BUG? i restart the UI and my PC but its still not available, the extensions are installed but are not visible.... Someone know what happen??
its not appearing even though the extension is installed
Would you be so kind and do a top 5 stable diffusion extension video soon?
Is it possible to create anime style characters with this? I tried using the blender 3d model for quite a few Anime models and it always tried to make the face Look realistic… is it the same for this method aswell or is it also trained for anime?
It sounds like you need to change your model to an anime based one and add more positive and negative prompts.
@@stinkyjutsu i already did that and it still tries to generate a realistic face
That's interesting, sorry that I wasn't any help.
@@prace2k If you're still working on this, that's when you tinker with the prompt to emphasize the anime style.
Witch extension is the icons that are in your prompt area?
I noticed the camera position is lower if use open pose and depth together, any solution on this?
Make sure the resolution for the images in control net match the resolution your trying to generate
4:22 Maybe it took in consideration, maybe she's doing the sign behind her back, we don't know
posex isnt showing that 3D windows, took me 2 hours to find a older version that works oO
Downloaded just fine but no model is there, just a blank box.
when I do the tutorial, the hands are dark in color
Apparently this does not work anymore, you should update..
My only problem with these open pose models is that they default to full body and what I really want is waist up
You should be able to drag the lower parts out of Frame and that should be all you need to do :)
Personally I am Going to stick with the Blender Addon I Am More used to it.
🙏
Sorry but 3d mode doesn't work
Juste need to wait 1 month for the same with feet 😂
If someone knows how to make a depth map picture of any normal picture??
He covered that in on of his previous videos. I think it's the one that says "This Is Huge" in the thumbnail and has two girls with their hands in front of their face
BROOOOOOOOOOO
Is there a way to put an image as background to try match the pose manually with the plugin?
Posex collapser is empty for me.
What about openoutpaint
not terrible not great tbh
@@Aitrepreneur
Well, we trust your point ❤️👍
hi
it produces only stickman :D where did i go wrong i wonder
Everyone is complaining in the recent comments and yet you ignore them. What about warning us in the title that the extension is outdated and broken instead of ruining people's stable diffusion? Dislike.
I noticed less joy in your voice, is everything okay?
PoSEX 🤣 has an interesting meaning in German
Das ist wirklich unlustig
Which is???
@@Aitrepreneur A-S e X
We need the same for feet, they are drawn just as bad as the hands by ai.
You can upload your own feet library for that or ask if other people did it
Anybody working on feet?
I tried but SD still has issues with feet.
SD thought the control net images was the feet from the bottom instead of from the top.