THIS IS CRAZY!!! The perfect poses made easy! Multi ControlNet, PoseX, Depth Lib, Stable Diffusion

[@thaido1750]

Ok no joke, having a low-end graphics card (RTX3050), i can't run blender fast enough to pose, however, the method you provided using the pose website was a savior, thank you so much for your help, and i gave the vid a like

[@LevendeStreg]

Thank you so much, Thai Do - I really appreciate it and glad I could help. Thanks for watching and commenting!

[@eddiedixon1356]

Awesome! Thank you!

[@thaido1750]

@Capitan Nerevar well yea when comparing it to 3060 and 3070

[@LevendeStreg]

@Capitan Nerevar yes you at the very leasy need RTX 3090 TI to run the extension and Blender. And that card is almost impossible to find.

[@thaido1750]

@Capitan Nerevar welp i beg to differ, i just upgraded to RX 6700 XT 12gb (Equivalent of 3070) and the margin of difference is vast, and we're talking about rendering here, you can't render fast or detail when you only have 4Gb of VRAM, you need at least 8Gb of VRAM in order to render something, and from a guy who used to owned RTX 3050 4Gb, it's not gonna do any good in the long run, and another short story is the price of RX 6700 XT from where i'm from is the similar as RTX 3050 8gb version, so why shouldn't i pick the performance similar to 3070 with the price of a 3050? I'm on a budget here so i only grab what i can, and now i can use Blender without worries that my graphics card is going to combust, besides, Blender do support AMD so that's a bonus

[@energeiai99]

cool - thanks

[@LevendeStreg]

You're welcome. Thanks for watching.

[@Aitrepreneur]

Great Workflow!

[@LevendeStreg]

Thank you kindly. I’m awestruck 🙌

[@traida111]

god damn you are made for tutorials lady, a breath of fresh air to find someone with passion and knowledge

[@LevendeStreg]

Thank you kindly, @traida111, I really appreciate it! ControlNet doesn't seem to work on SDXL though, just if you were wondering. Thanks for watching and commenting!🙌

[@traida111]

@@LevendeStreg no worries :) i am using automatic1111, i find for me the colour stick men dont translate difficult poses. Tonight i plan to try using a reference image of photo with pose already in position, then i might try the canny line art, i will try select the processor then the button there can output a preview, so i can check it and adjust the stength until the canny lines are suitable. If i can get it accurate and easy process i will be happy. I tried milti control net, with openpose for body and depth map for hands and feet but i couldnt get it to work well

[@edsonjr-dev]

💜💜💜

[@slashkeyAI]

Happy to have found your channel.

[@LevendeStreg]

Thank you kindly. I really appreciate it. And thank you for watching and commenting 🙌

[@munyunu]

you are such a nice person thank you for the video

[@LevendeStreg]

Thank you kindly, Min Yunu🙌 and thanks for watching. I really appreciate it!

[@ujjvalw2684]

That's something new . An Ai Artist .

[@LevendeStreg]

Hahaha. Yup precisely! 🙌

[@JohnVanderbeck]

The real problem with OpenPose in ConttrolNet is that it doesn't project in 3d properly. There is no perspective in the "Bones" that it uses so it has no way of knowing what depth they are in. It is a significant disadvantage, despite how amazing it is.

[@LevendeStreg]

Yes that is so well put. And that’s why I suggest the depth model in stead. Thank you for commenting and being articulate and to the point. I should have pointed that out more clearly in the video🙏

[@MultiHeheboy]

Stable diffusion itself doesn't understand or not trained like this. So it's possible to input 3d information.

[@nicolasgabriel3402]

Have you found a way around it? I was thinking what is the best workflow to pose with some perspective or depth and I found this video and no so much more...

[@thechosenone729]

This is how i imagine smart artist ... instead of constant complaining about how it's going to take your job taking it and use it to your own advantage to make it even better. Good video btw.

[@LevendeStreg]

Thank you kindly Peter. I really appreciate it. And thanks for watching. So glad you like the video🙌

[@Smashachu]

Huh.. I've never been much of an artist or really ever enjoyed drawing beyond trying it and being disappointed. However.. I love stable diffusion and i love everything about being able to take whoever idea i have in my head and transposing it on paper with with speech. Maybe i'm not a real artist but this.. has become an obsession

[@LevendeStreg]

Hahahha that's wonderful. And yes making images with AI is highly addictive😜... and what is a real artist anyway? It's so great to hear that you found your inner creative!

[@erikdong]

brilliant video. thank you for sharing!

[@LevendeStreg]

Thank you kindly. I really appreciate it. And thanks for watching.🙌

[@peacefusion]

you know shes crazy good when shes wearing a hand brace.

[@LevendeStreg]

Hahahahaha... It's a glove for my wacom cintiq pro 32 inch. Otherwise my hands sticks to the screen and the ruins the flow when I'm drawing.

[@HieuTran-rl2qw]

tks you alot for amazing tools!!!

[@LevendeStreg]

You’re very welcome! Thanks for watching!

[@n8wn8wn8w]

You are a generous soul. Thank you❤

[@LevendeStreg]

Thank you kindly - and thank you for watching. I really appreciate it 🙏

[@thomasmann4536]

I love how excited people get about being able to pose inside Stable Diffusion and not having to use Blender. As a Technical Animator (the person who creates the rigs that allow you to move a 3d model around) I have no problem with using Blender and I just find the PoseX UI disgusting :D

[@LevendeStreg]

Haha. Thanks for commenting and watching. I totally get you point of view. I am not a fan of openpose and not of PoseX either. I prefer the canny and depth model😜 And ottally cool with a job where you rig characters🙌

[@ramiropiedrabuena3892]

Design Doll works fine.

[@LevendeStreg]

Thanks for the recommendation for Design Doll. Good to know!🙌

[@aaronhhill]

Being a HUGE fan of Aitrepreneur I have also subscribed to your channel. Glad I ran across it! Thanks for your work.

[@LevendeStreg]

Thank you kindly - I really appreciate it. Thanks for subscribing and watching!🙌

[@SaintMatthieuSimard]

As the US copyright office decided that they would not register copyrights on unique material generated with AI, I decided I would not use it... But that's still interesting to learn about it. It's not so different than applying a sephia filter on a photo on instagram in terms of "it's still your own". They should review their position using common sense.

[@LevendeStreg]

I totally agree, Matthieu. And I think that it will change over time, when people understand the process of working with it.

[@PADATWO]

I have just discovered your channel which seems very helpful. But ineed to improve a few my english ! Thanks for sharing 😊

[@LevendeStreg]

Thank you so much for watching and commenting. Cheers🙌

[@Neblina1985]

amazing video! i always love the way of your type of explanations

[@mariaprohazka2730]

Thank you kindly, José. And thank you for commenting and watching🙌

[@LevendeStreg]

Thank you kindly, José. And sorry. I answered that one with my private account😬

[@dodd15]

Your workflow is very clean and polished, best one I have seen so far.

[@LevendeStreg]

Thank you kindly. I really appreciate it. And thanks for watching🙌

[@dougmaisner]

cool stuff, i'm going to try these steps!

[@LevendeStreg]

Thank you kindly. Hope it works well for you. And thank you for watching 🙌

[@charlesmartel3995]

Great news!

[@LevendeStreg]

Thank you kindly. And yes. Great news! 🥳

[@lsycxyj]

Still hard to get the hands fixed. Is there any models with detailed hands?

[@LevendeStreg]

I’ll try to look in on it. And then I’ll get back to you🙌

[@AndysTV]

Is it possible to generate a character but also its decomposed body parts? How do I setup the prompt? Or base image?

[@LevendeStreg]

Yes, that would be possible. That's a great idea for a video! Thank you for bringing my attention to it. You could use open pose for that. And just zoom in on the different body parts.

[@zimnelredoran9985]

Hi, thanks for the great video, Controlnet is beyond fascinating :) Just a question, are the steps we see already done (like getting a pose done in the third tab of multicontrolnet) explained in another video? Another question, how do you send the just the hands to the controlnet tab? Thanks :)

[@LevendeStreg]

Hi there and thanks for your comment and question. I think I’ll be doing a video where I explain it in a little more detail. But basically you download the png file to your computer and then upload it to the tab. 🙌

[@zimnelredoran9985]

@@LevendeStreg Thanks for the fast response : )

[@sirusazadi2349]

Great video! Are these apps installable on a Mac?

[@LevendeStreg]

Thank you kindly, @Sirus. They're not really apps, but code bits. And yes you can install Stable Diffusion locally on a Mac. But there are some of the code - like the Dreambooth extension that doesn't work on a mac. That's why I use RunDiffusion for running the code on Cloud GPU. Remember to use the promo code (levendestreg15) for RunDiffusion if you sign up for Creator's Club. That will get you 15% discount.

[@sirusazadi2349]

@@LevendeStreg thank you, really learning a lot from you

[@danwood4171]

I came to find out how to use posex and saw no tutorial on it. I came create the openpose thing, and tell it to send it to control net and enable control net but I'm not getting my person like the pose. I've used openpose directly where the pose is created from an input image and that works but I can't get a posex created "skeleton" to work. But I saw a lot of hints that great power is available with "other" things which is nice.

[@LevendeStreg]

Thank you for your comment, Dan. You have to install the ContralNet extension, Depth Library and maybe the PoseX. Then download the ControlNet models and install them. And then in your Automatic1111 you go to settings, and then down under controlnet you switch on multi-function. This way you can add more ControlNet tabs to your images. It's a complex thing to get working. I'll see if I can do a new video on it.

[@danwood4171]

@@LevendeStreg Thanks for the reply. I just needed to realize preprocessor needs to be set to None. The way I now explain it to others is: OLD WAY: Two steps: 1. CN image preprossesing -> depth map, stick figure, etc. 2. preprocessed result -> image POSEX WAY: 1. You pose the stick figure and then generate image without preprocessing. Of course, a demo is better than words. FYI, the overnight set of A1111 changes has broke posex and perhaps other things. The bugs have already been reported on github.

[@LevendeStreg]

@@danwood4171 Thank you for that Dan. Really appreciate you breaking it down for others to follow - I probably didn't make that clear. And yes, A1111 often breaks😬 - I'm using RunDiffusion though, and they have it fixed.

[@geoatherton5214]

Love your videos Maria, thank you for all the hard work you put into making them. I agree with all your same conclusions like ‘I only really want to use AI tools if they are faster than just drawing it.’ I have a graphic novel project as a consistent hobby, so I appreciate hearing your perspective as you look at AI art tools through the lens of a comic book artist. If you’ve clocked the hours learning to draw figurative anatomy, that is faster and generally more fun/loose/enjoyable than painstakingly moving joints around in blender or open pose, which is way more trial and error and technical troubleshooting than the flow state of just drawing. What I love about AI art right now is drawing a loose sketch with a basic color palette and maybe 2 tone shading, then feeding that to SD or Midjourney and just having it layer on painterly detail in the blink of an eye (I’ve got some demos of this on my own channel). Huge time saver. This latest raft of Control Net AI tools are amazing in that they empower people who haven’t had all those drawing practice hours to express themselves and tell their own stories, but I’m just as excited for all the possibilities it unlocks for practiced illustrators to supercharge their existing powers and output volume while staying ‘in the flow.’ Keep ‘em coming, salutations from Seattle and looking forward to your next one. :)

[@LevendeStreg]

Oh wow. Thank you for watching, commenting and sharing your knowledge. I love the workflow you show in your videos. That is precisely what I’m talking about. Great use of AI. Can I connect with you on Discord? Would love to share thought on some of the processes…

[@geoatherton5214]

@@LevendeStreg Definitely! Just sent a DM to your Instagram. Thanks again 🤘🖖🎨

[@LevendeStreg]

@@geoatherton5214 Thank you. Hooked up with you on Discord🙏

[@pladselsker8340]

I love the energy you give. Great video!

[@LevendeStreg]

Thank you so much. I really appreciate it 🙌

[@pritamsarkar2385]

how can i get consistent results? i.e same character wearing same outfit but different poses of my choice without any crazy/ugly deformation?...every time different results is frustrating me

[@LevendeStreg]

Yeah, that's that though part. The technology is not quite there yet. It's difficult to get consistent outfits - without the deformation. Try doing long negative promtps. That sometimes works. Good luck.🙌

[@chariots8x230]

It’s really impressive how things are progressing with AI art technology. I like that you were able to pose your own custom character. Although the results are pretty decent, the problem with 3D a lot of the time is that it produces poses that look stiff. If the AI copies these poses exactly as is, then it will also produce stiff-looking characters. It would be cool if there was a setting to maintain the overall pose, but modify it very slightly to get rid of the stiffness to create a more natural-looking pose. I’m excited to see how this technology will evolve in order to create whole scenes. For comics, I’m waiting for features like: 1 - Natural looking poses (no more unintended stiffness) 2 - Posing multiple of my custom characters together in difficult poses (e.g., hugging each other, holding hands) 3 - Changing the facial expressions of each of my custom characters in the scene 4 - Using some nonhuman custom characters such as animals, creatures, and anthropomorphic characters 5 - Using custom backgrounds that can be shown from different angles, lighting conditions, time of day, weather conditions, etc. 6 - Combining my custom characters with a custom background in order to create a full scene Thank you for sharing this video 😊

[@LevendeStreg]

Thank you so much for your comment @chariots8x - I really appreciate the time you took to put together your comment. And I totally agree with you. There is a lot of stiffness to the 3D poses. I found that Stable Diffusion’s “translation” of the pose is not as stiff. At least not if you turn down the weight of the ControlNet pose. And as you say, the development of this is impressive. And in a couple of month it will be so much better. This is all so exciting to follow and learn. 🙌

[@chariots8x230]

@@LevendeStreg Yes, it is exciting to see how this technology develops and the new things it can do. Maybe someday it will become suitable for complicated projects, like comics. ☺️ Hopefully, soon we’ll be able to create full scenes with multiple consistent characters & backgrounds. The ability to copy a pose and apply it to a single character is a great development. Now if we can apply interactive poses to multiple characters in a scene, and also give each character a different facial expression, that’d be great.

[@LevendeStreg]

@@chariots8x230 I think that is already doable with inpaiting. But you're right. Right now it's still easier / faster to do the drawing instead. But in a couple of month... I think this will move fast.🙌

[@bumstudios8817]

I think they can already do that.. it’s just working out the bugs for now. They are able to replicate a image but keep the same style.. I have seen a video where they just used an image sequence for the animation and the ai did the work for the style look.. then put it together after in a video editor to make it a movie again..

[@bumstudios8817]

P.s which means you could use an actual video or movie of real people moving eliminating the cgi animation stiffness

[@obsolete.camper]

Thanks a lot, I learn a lot from your videos. Your explanation is clear and easy to follow and please keep it up, I would like to learn more about techniques you use to make SD more effective to use. Regarding this tutorial, can you try to explain again STEP 5 when you mention about the "different" tabs in the Controlnet section. Are you trying to say turn on the multiple Controlnet tab features?

[@LevendeStreg]

Thank you so much. And yes, I should have explained that more clearly. Yes I Multi ControlNet switches on in the settings. But I actually meant that you also need to enable the different ControlNet tabs.

[@emilianovb]

Awesome video! The algorithm recommended this to me at the perfect time! I spent the last two days learning controlnet and improving my txt to image prompts. Quick question: Would this method work on multiple subjects ? (two or three 'people' in an image?) Thanks!

[@LevendeStreg]

Haven’t tried that out yet. I think I would add them in one image though. So combine three images into one. But yeah, I guess you could do it the other way. In my experience, though, 3 ControlNet tabs I tops. They mess up the code - it becomes wonky and unstable - and the output becomes a bit weird sometimes.

[@LevendeStreg]

And I’m glad you like the video. Thank you🙌

[@alejandromarello6907]

Quiero hacer un trailer de 1 minuto de mi guión cinematográfico. Que programa o software me recomienda?!

[@LevendeStreg]

Well that sounds like a great idea. But a trailer can be quite a lot of work. You need to create a lot of footage. And even though AI exists - there is no easy solution. I would probably look into Next-Gen2. But there's a long waiting list there. And I don't know which style you want to do. You can do a lot of cool stuff with Stable Diffusion - but you need to render it frame-by-frame - but with a setu so you render many frames at a time.

[@alejandromarello6907]

@@LevendeStreg thanks. i'll keep looking into it then

[@djanevski]

This is awesome! Thank you. Great quality video as well! Keep it going ! I will try and let you know

[@LevendeStreg]

Thank you kindly , Danny. I really appreciate it 🙌

[@alisoncristinazertuchelope7332]

Hello, good video, but why don't I get the 3 net control windows? The ones that are Control Model - 0, etc. I have the net control version e1885108 (Wed Apr 12 03:24:32 2023) and the posex one (Sat Apr 1 12:28:38 2023) which is the new one.

[@LevendeStreg]

Hi there and thank you kindly🙌 - to get Multi ControlNet go to Automatic1111 >>settings - and look for ControlNet. Then slide the slider up to 3 controlnet tabs. Hope it works out for you.

[@dongivafoc4493]

Hi, very insightful video. I am struggling to get consistent results on the poses. I have trained a model on a real person (me) and I'm generating a somehow similar appearance to the real person but you can still see that it's 2 very different people. it's very difficult to get the face and the poses consistently. Either the face will be acceptable or the pose but very difficult to get both of them. So I will be trying your solution, but I was wondering, since it's been 4 months now, if there was other tools I could use to get consistent results. I'm using stable diffusion and the installation of depth lib is making it crash. Is this something that doesn't work anymore or just an issue on my side?

[@LevendeStreg]

Depth library is difficult and crashes a lot. That’s why I use RunDiffusion for it. You need multi ControlNet - and then The Roop extension. I did a video on that too. And maybe use the new upscaler. They’re cool and you can sort of upscale like a puzzle. I’m gonna do a video on it soon. Hope it works out for you! 🙌

[@Grapegum]

Who messed up the timeline to the point Sarah Connor is an AI artist? Jokes aside, this is one of my favourite AI channels ever, keep up the good work!

[@LevendeStreg]

Hahaha. Thank you so much. I really appreciate it 🙏 thanks for watching!

[@kinlih289]

helpful but chaotic

[@LevendeStreg]

Glad you liked it and thanks for watching. Sorry for it being a bit chaotic. My brain is a bit chaotic!

[@kaiandree]

Cool! Thanks! As being an illustrator, I wonder wether it wouldn’t the easiest way for me to just scribble the character.. can you please make a video about that way to produce photolike images?

[@LevendeStreg]

Thank you kindly. I really appreciate it! And yes I will try to make a video on that. Thank you for your request!

[@mitamoto3399]

Love this, thank you for so many amazing tips! However, I did find myself getting a little sidetracked at times as you jumped from one topic to another pretty quickly. It would be really helpful if you could provide a more step-by-step tutorial on how to insert images into Posemy art instead of just showing screenshots of how to do it.

[@LevendeStreg]

Thank you kindly for you comment and thank you for watching. I have an upcoming video on what you request. An in depth video on posemy.art 🙌

[@mauricioc.almeida2482]

I would like to express my sincere gratitude for this amazing video tutorial. Your help was invaluable and allowed me to make significant progress in my hobby as AI creator design. I was very impressed with the clarity and quality of your tutorial, which made it easy for me to understand concepts that were previously confusing. Thank you again for sharing your knowledge and skill with us. Your generosity is greatly appreciated, and I look forward to continue learning from you in the future.

[@LevendeStreg]

Thank you kindly, Mauricio. I really appreciate it 🙌 thank you for watching and taking the time to comment.

[@user-qr4jf4tv2x]

i can't believe its not blender *slogan

[@LevendeStreg]

Hahaha.... Right! It's pretty cool!

[@ellenripley4837]

I have another inquiry. does this method works with images you generate elsewhere like midjourney? I don't like the results stable diffusion gives me when it comes to style and I have nailed a Gouache cartoons style on midjourney that looks like something I would do but I was wondering how good this is with a more stylized illustration.

[@LevendeStreg]

Thank you for your question. To my knowledge you cannot achieve the same results with Midjourney, no. Only Stable Diffusion can be controlled like that.

[@hogyodoshin]

Hi, I am on Windows, Have installed posex on 2 different PC but the 3D pane and controls are not there: any clue what happens ?

[@LevendeStreg]

Don’t know sadly. I’m on a Mac - so I haven’t had the best results with local installs. Have you checked the settings and updated?

[@4538304544]

same problem here, any solutions?

[@LevendeStreg]

No, sorry to say so@@4538304544 . It's probably that the code is broken yet again. Or that your graphics card isn't big enough. You need a Nvidia (pretty big card) to get 3D to work. I'd recommend checking for updates. Delete the old install and do another install. Sometimes the code is updated, but that's ends up causing bugs in many of the extensions. It could also be a A1111 bug (that happens very frequently). Maybe try out RunDiffusion.com (they have all the newest installs on that - even comfu UI). Remember to use the promo code levendestreg15 to get 15% off on Creator's Club (you need that to run SD with Multi ControlNet). Hope it works out for you!🙌

[@darth_sidious_sheev_palpatine]

Can this method work with two actors? If I wanted to always have two people fist bumping on a single promt, or have two actors crossing light sabers in a fight image?

[@LevendeStreg]

Yes, but you have to create an input photo for ControlNet using an image with two people.

[@Alemmo-x5l]

Hi, I installed everything but the "depth library" tab doesn't appear. How should I proceed about this?

[@djfremen]

Looks like you're running on OS X. So if it's apple silicone (m1, m2) then you are forced to use RunDiffusion only you have the earlier intel version / eGPU, correct? Somehow I think these heavy extensions will not work leveraging CPU and RAM only...

[@LevendeStreg]

Correct. I’m running on Mac m1. And you’re right - on RunDiffusion I’m using the Large server. The medium cannot run them - or sometimes they’re buggy at least. I am looking into getting a pc with a Nvidia (maybe rtx 4070ti, but not sure yet. This is all moving so fast. Who knows what the requirements will be in a couple of months… so right now RunDiffusion works well for that.

[@djfremen]

@@LevendeStreg good idea. I've seen people repurpose old crypto mining rigs from alienware aurora r12 with the rtx 3090. Seems like a cheaper way to go than building a whole new PC.

[@krz9000]

Get a 4090...if you use this to make money there is really no reason to not want to have the best available. Especiwlly with something that is at the core of your process.

[@djfremen]

@@krz9000 I still think putting everything inside a spreadsheet and calculating the cost vs buying hours on run-diffusion. R12 3090 with CPU/RAM 1k, 4090 new 1.8k for only the card!

[@LevendeStreg]

@@krz9000 hahahaha… yeah you’re right. But they’re so expensive 🤯

[@CadyCadwell]

max res i got while using controlnet on 6gb vram is just 256, can't even do 512. any workaround?

[@LevendeStreg]

Nope sorry, it's crazy heavy on your computer and GPU to do ControlNet.

[@neilrhodes3879]

The screne gigging up and down is a nightmare to follow and not sure if you are trying to draw attention to something or if you are just wiggling your mouse, im a noob with this so could not understand what i was looking at as your SD screen is moving upo and down all the time

[@LevendeStreg]

Sorry for the inconvenience for you. I'm learning all the time too, both with AI and with making learning videos. Thanks for watching.

[@MrFedemoral]

Hi Levende!, i have to ask again, i cant get preview frames and progress bar working in GUI. I tried everything reasonable. I checked all possible settings imao. Obviously applying and reloading GUI sooo many times. image finally appear when render is complete, but i cant see anything in the process, also IM DESESPERATED as i tried everything. Any idea why it doesnt work?

[@LevendeStreg]

Hi Morality. I’m not sure I can help. Maybe it’s a settings thing - or maybe your install is broken. Working with ControlNet takes a lot of gpu up - and if your graphics card is not big enough it will produce black or weird frames.

[@hagardeviking9314]

can I export to SL?

[@LevendeStreg]

What do you mean with SL?

[@hogyodoshin]

Does not work anymore cause of the last stable diffusion update update..

[@LevendeStreg]

Well, as far as I know it works on RunDiffusion. There’s always a bit of trouble with the updates😅

[@SPYBGToolkit]

I'll make sure he sees the video ;)

[@LevendeStreg]

Hehehe - thanks. I’d be honored! I’m a hard core fan 🙏

[@SPYBGToolkit]

@@LevendeStreg by the way here is something that you may be interested in, since you do art as well :) give it a try let me know does it work for you kzbin.info/www/bejne/fmWUY41sg9tsrKM

[@LevendeStreg]

Whoops - answered you with my person YT. Sorry about that. The video you shared is great - and a great idea too. Thank you so much for sharing this with me!🙌 by the way - I love this video you did: kzbin.info/www/bejne/n6TQiKGfnrlrm9k - I'm gonna try do show some of the same. But done in a different manner. Love your videos! And now subscribed to you!

[@nvadito7579]

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?

[@LevendeStreg]

Sorry to hear it. Sometimes the multi ControlNet just freezes up for me too. Normally when I don’t have enough GPU that can happen.

[@roberthintz4017]

This should be extended to animals.

[@LevendeStreg]

You’re right. That might work too.

[@crashdummyglory]

the video should start from the 9:18 mark tbh.

[@沙漏下的足跡]

我找時間來學習這個功能。

[@LevendeStreg]

I'm glad you learned from this. Thank you for watching 🙌

[@josevieira3367]

:)

[@LevendeStreg]

Thanks José! And thanks for watching!

[@ellenripley4837]

isn't it better to get one of those really good dummy dolls with ball jointed articulations to create poses, then take a picture then edit it in photoshop and voila! I find the rigged 3d doll so annoying to use.

[@LevendeStreg]

Yeah, I agree. I actually have some small dolls too… that you can use. And I’ve wanted to do an episode on that for some time now. So thanks for this question! 🙌

[@johnsmith4402]

If so complicted like this, what is the point to use AI.

[@LevendeStreg]

That is a really great question and good point. I think AI has been harnessed by huge companies like Disney (and thereby Pixar too) already. And with great skill they can produce animated videos with it already. But for small users like me and you I’m afraid the gain of using AI for image creation is not that big. The computers and graphic cards need to be huge. And we cannot - as small users - save time using it. It’s complicated to use, the code breaks all the time and it’s still way easier to do by hand most of the time…

[@genin69]

at around 8:07 is it possible to rather have a small floating head instead of that massive 16:9 video overlay on the actual workspace. i feel it covers a lot of what needs to be seen and is also distracting..

[@LevendeStreg]

Thank you for your comment. I’ll see what I can do in future videos! 🙌

[@eod9910]

do not do this on your computer locally because you will fuck up your computer and have to reinstall stable diffusion all over again you will lose all of the extensions that you installed

[@LevendeStreg]

Yup, that’s why I use RunDiffusion for it. This way all extensions are valid and stable! 🙌

[@TeleMA50c]

This video contains a lot of good information but I found it hard to follow. You repeat yourself a lot and don't walk through the process in a linear way but rather a meandering, out of order sort of way. Not to sound ungrateful but if you were able to edit the video to be much clearer and more linear it would improve it tremendously.

[@LevendeStreg]

Hi there - first of all thanks for watching and commenting. I really appreciate it 🙌 secondly, you are quite right. I’m a creative spirit and my brain jumps around a lot and it’s a complex subject about ControlNet. I’m trying all the time to improve, so I am already trying to follow a more linear approach in my videos. 👍

[@St.MichaelsXMother]

are there any Like FREE LIGIT Lineart converting images on google you can Link me?

[@LevendeStreg]

What a great question. The only one I was able to find with a quick search was this one: tech-lagoon.com/imagechef/en/image-to-edge.html - but I'll look into it and get it into one of my next videos. Cheers 🙌

[@LevendeStreg]

But Adobe Illustrator is really the best for that. I use that feature all the time!

[@thevoid6756]

I am surprised an illustrator like yourself has to use clunky tools like openpose or learn basic 3D tools to get AI to pose a character correctly. I would imagine it shouldn't be too dificult to basically let artists freehand a sketch with img2img and add text to refine the prompt.

[@LevendeStreg]

That's a great comment. And you're right. But - I'm testing out what gives the best results. And I have to admit the 3D poses actually gives me better results right now. I still haven't gotten my process down to what I want it to be. In the beginning I thought that I would get the best results with what you describe. Img2img and ControlNet. But to my great surprise that wasn't the case. I need to have at least the colors in place too on my sketch for that to work. And also I find that I get the best outlines on my comic book images if I use 3D poses instead of sketches right now. But again - I need to play around some more with ControlNet, because I'm actually really keen on using the Scribble model. Still I'm getting better results with Canny and a 3d model image right now.

[@OnionCurry]

Filtering stolen art by similar poses! So cool!

[@LevendeStreg]

I'm not sure I understand what you mean. How is it filtering stolen art? I have a great conversation and colab with posemy.art (if that's what you're alluding to). Otherwise can't see how it's stolen? I'm an artist myself, I do most of my work by hand, but I also use AI where I can. I'm trying to tweak my work process and to help other artists do the same.

[@kurwa_bmw]

11 minutes ouf of 12 are just about promos and the rest is full crap

[@LevendeStreg]

Thank you for leaving a comment and thanks for watching.

[@speeddemonau6802]

I couldnt watch to the end as I found it tedious, especially the excessive banter. I also found it lazy that instead of filming and editing the complete process, you tacked on something at the end. Given you announced the 'additional content at the end' at all stages of the video, it seemed to be a design choice to force watching to the end of the video. Provide content worth watching and I will watch every second...

[@LevendeStreg]

Thank you for your feedback. It takes forever for me to learn all the new stuff AND do these videos - even though it doesn't seem so for you. My brain jumps around a lot - and I cannot remember a script. So that's why the excessive talk is there. And I cannot cut it out - because part of the background video (with Auto1111 UI) will be missing. A video like this one I probably put at least 15 hours into with research and filming and editing. It's very hard to do as I run a small visual agency full time. And this is just something I do for interest and for teaching others - I don't really make money on the videos as such (not even to cover costs). But thank you for watching.

[@kiv1198]

AI 'Creators, Designer, Engineers' what a bunch of entitled words just to avoid saying "AI User". Just accept what you are and stop using real effort titles.

[@LevendeStreg]

Well Thank you for your comment. AI user is only very little of what I do. I’m an animator and illustrator - and I run a small visual agency. So that’s actually what I spent most of my time on - not AI. It’s a very small part of my workflow.

[@ChrisVirgilio]

Over 3 and a half minutes in before you finally introduce yourself and tell us why we should even bother listening to you? Gosh.

[@Ia960-s4l]

No artist you are just crawling other’s art and printing it randomly. Ok, theft.

[@LevendeStreg]

Thank you for your comment. And nope I’m NOT - that’s not how AI works. I use AI to speed up my creative process at times. And at times it easier and quicker to just draw everything myself. You would know if you’d spend the time learning about AI diffusion models and if you’d seen more of my videos. I started out as an illustrator and animator. I still do lots of live graphic recordings and company visualizations, hand drawn comics and animation. And I continue to learn about new and evolving stuff that comes to the creative scene and helps visual artists. Hope that you have a great Sunday 🙌