Just watched this one and it blew my mind - again. I'm still learning to apply the techniques you explained in previous videos! There's a gold mine of info in everything you've released. Thanks so much for sharing your insights.
Como un desarrollador de cómics independiente que está intentando hacer cositas con inteligencia artificial y básicamente tiene que colocar fondos blancos para cada uno de los pinches paneles, sólo puedo decir... ERES EL PUTO AMO TÍO
Hahahahaha gracias! Me alegro de que ayude! El proximo vid es de meter los personajes en los escenarios e interactuar con estos. Se te ocurre alguna otra cosa que pueda apuntarme en la lista de videos a hacer?
@@Not4Talent_AI Lo único adicional que se me puede ocurrir es cualquier otra cosa que tenga que ver con desarrollo de cómics, léase, cómo cambiar ropas para que, por decir, un kimono que tenga un estampado distintivo, lo tenga en todos los dibujos que hagamos, o cómo cambiar de estilo sin cambiar la composición, cosas así
this could have application far beyond just 3d rendering. Conversion of the image into 3d object is effectively object detection algorithm for self driving vehicle. I could see stable diffusion leak into self driving vehicle sector in the future.
that's thinking so far ahead my car crashed twice just thinking bout it. Sounds super interesting tho, even if I have absolutely no idea how hahahhahha
@@Not4Talent_AICurrently in the self driving tech they mainly use object detection from the camera frame. Then attempt to project detection coordinate onto a xyz coordinate frame. Then the car operate from xyz coordinate frame. It has similar issue of temporal and spatial inconsistency that AI video have. IE: if a car went behind truck, the algorithm doesnt know it existed anymore. There is a lot of back end code that try to stitch together detection across multiple frame to maintain object temporal consistency. This workflow of image to 3d blender effectively convert image or video data directly into 3d object. So we don't have to stitch together data to provide temporal consistency. A lot algorithm is share between the AI 3D generation space and object mapping space. Since gen AI is converting from coordinate (or lower dimension promt) to 3d object, and 3d object detection is the opposite workflow, converting from picture to 3d space.
oh wow that's extremely insightful and cool ngl. Wasnt expecting to learn about this today hahahaha ty for the detailed explanation!! I'm not gonna say if you are right or not cuz I have 0fking idea, but you seem very very knowledgeable on the topic. So amma trust you hahaha Would love to see if this goes somewhere ngl@@keenheat3335
@@keenheat3335 nice input htere. i knew convolution neural network was used as part of image identification process so it make sense as it is at the root of the caption process of images in SD. i don't dwelve into that but i see te point where a 3D depth combined is much more recognizable than a flat 2D shape. good luck on your work, i could have just been a bad idea from the start but history have shown that discovery are made by accidents sometimes.
Excellent video my dude, gonna have to bookmark this one for later, because I'll have a use for this knowledge in the future. Any checkpoints (anime and/or realistic) in particular you'd recommend for environments?
thank you so much!! I usually just use bout 3 or 4 models: realistic 1.5: EpicRealism realistic SDXL: ColossusProject xl anime 1.5: Anylora or ivory AnimeSDXL: (haven really used this much, probably better ones are out there) animagine xl
sorry i'm late to the party, i had to motivate myself. that's a lot of decation for the niche audience that would be willing to use it. but now thanks to you i managed to understand what i missed from active motion picture instruction on depth map inside blender ! and there is much more than what i was expecting. i'll be waiting your next video. i would definitly need that. My 3D waifu never managed to drink the beer in my Lora Samples !
thanks!! Idk if my next vid will touch much 3D, didnt find it necesary to use it as a way to integrate characters. Maybe i'll teach my version of clip drop relight at home, but I'll see. I still need more blender practice to make tutorials on it hahahahaha. What was it that you where expecting to use it for btw? Just in case I can add it or I can make a separate short/video on it
@@Not4Talent_AI the ai feels you that we see often ? It remain. Making it appear dimer with a custom character/style is a nice way to prevent part of it. However the next step is to control the environnement, the object interaction and the lightning of each scene and make it fit acurately. (Your use of reference/IP adapter was a great point toward that). Changing character shape and shade with the coherent background is Key to make a great masterpiece. One that i would put as desktop wallpaper without anyone noticing the ai things. (We are far away still :) )
Yass! for now I'll focus on adding the character, and trying my best to make it integrated into the scene, with proper lighting, prespective, scale, color... etc As well as the character being able to sit, grab, break... parts of it. @@lefourbe5596
Great tutorial! Your clear instructions and simple approach make tricky concepts easy to grasp. Excited to test these techniques in my projects. Keep it up! 👍
Holy shit, feels like that video cam after my feedback that I can't apply that knowledge to the game development ^^ Thanks for showing how to create my own HDR textures with SD!! That helped a lot! If only I'd stop struggling with the basics of Lora and In-/Outpainting, then u would be able to apply this to my job. And also thanks for showing us the 3D modelling AI! I hope there may exist some locally install-able AIs in the future, just like SD 🙏 And I appreciate so much that you give us free access to your videos! They are definitely high quality and contain a lot of effort. I wish I could donor smt. in return, but as a student in game developmeny my pockets are kinda empty ..
happy to hear it helps!!! If you have any issues with inpainting or outpainting, just ask! After posting a video I usually help around in the discord if I can, so it's a good time to ask any questions hahaha Also, no need to donate at all, comments like this already make it all worth it. As long as the vids I put out help yall, thats more than enough! as always, thank you so much for your kind words!
@@Not4Talent_AI This mindset is rare. You are such a good fellow. Thank you! You really don't mind if I ask you stuff via DM? (I mean I can trade it for Game Dev and Coding knowledge ;p)
Thanks for the mention! BTW, Euler - pronounced Oiler :) Also, I haven't even thought about using this for unique Skydomes for games :o should be interesting and also, "But sir, I want to type words and get a full 3D environment in seconds 😡"
Directly from text to image or with image to image without extra upscalers? In txt2img you could do lile 1024x2048 and then highres to 1.6 or so I think. In img to image similar sizes. If you want you can upscale with ultimate sd upscale or rezise with tiled diffusion. There you pretty much go 8k if you want
idk, ask him. If I knew I'd use it too hahahhaha (its most likely not even secret, he just has different settings than mine but I didnt ask for them xD)
I don't think it can. As a 3D modeler I have the skills to adapt to it and implement it as a part of my workflow. I'm hyped for it to save me time in the mundane aspects of the job, and hyped to learn a new tool that has never been explored before. If at some time it is good enough to replace me entirely, I have other skills to keep the job. so I'm confident and lucky enough to not have to worry. Ofc if my situation was different, I might not be as hyped.
thanks!! Tbh, not really hahhaahha the only portfolio I have and can show is probably like 4 years old... so I'd rather not share XD And the rest of my asset-focused portfolio is based on projects I've worked on and I'm unsure if I can legally show them. I have a 3D printed hollowknight figurine that I made which I could take a pic of I guess XD I mainly do lowpoly assets for games and web tho @@notDreadful
@@Not4Talent_AI You should give yourself more credit, besides your 3D portfolio is bound to be better than mine, which just consists of terrible 3d animations from like 2009.
I swear, everytime i have an idea that i don't know how to materialize, you deliver with an awesome video... can you read my thoughts ???
Hahahhahahaha tyty!!
I wish I could if they are full of cool ideas 😂😂😂
Workflows that combine Blender and Stable Diffusion is definitely something everyone should have their 👀 on. Love your work, keep going! 🎉
agree!! and thank you hahaha
Using Stable Diffusion to generate HDRIs for nice scene lighting in Blender is an area that's pretty exciting to me, personally.
Oh Yeah, this is 10000 going into my Unreal Pipeline.
Just watched this one and it blew my mind - again. I'm still learning to apply the techniques you explained in previous videos! There's a gold mine of info in everything you've released. Thanks so much for sharing your insights.
Ty so much!!! Hope it helps!
Cool thanks!
you don't need those random stock clips.. they're just distracting, like ads..
Hahahhahah I agree, I used them cuz my recording was blury 😂
Didn't even know this was possible, thanks! Gonna run through all these steps when I have some time!
Nice! Hope they help jahaha
Cant help but to amaze at how did anyone manage to figure all these out, well played dude im definetly learning blender with this ❤
hahahah most of it is just mixing existing ideas together tbh. but tyty!!
And please do, 3D is such a fun thing to learn
You really do drop the gems. This was a well put together tutorial. A lot of value to take in. Thanks mate.
Hahahahha thank you!!
Great tutorial!! It was the only video on consistent places I could find on TH-cam
tyty!! hope it helps!
Thanks you very very much buddy, it's the most detailed channcel about the stable diffusion i found in the intire youtube, thanks again man.
Happy to help!! Thank you for watching and leaving such a positive comment!
GOD CONTENT !
thank you very much!!!
Video highlight: 6:14
typical AI enjoyer moment
I honestly dont get it 😂 the depth map?
If it was boobs then Id say hell yeah, but idk if depth maps are it
Como un desarrollador de cómics independiente que está intentando hacer cositas con inteligencia artificial y básicamente tiene que colocar fondos blancos para cada uno de los pinches paneles, sólo puedo decir...
ERES EL PUTO AMO TÍO
Hahahahaha gracias! Me alegro de que ayude!
El proximo vid es de meter los personajes en los escenarios e interactuar con estos.
Se te ocurre alguna otra cosa que pueda apuntarme en la lista de videos a hacer?
@@Not4Talent_AI Lo único adicional que se me puede ocurrir es cualquier otra cosa que tenga que ver con desarrollo de cómics, léase, cómo cambiar ropas para que, por decir, un kimono que tenga un estampado distintivo, lo tenga en todos los dibujos que hagamos, o cómo cambiar de estilo sin cambiar la composición, cosas así
apuntado, gracias!
@@Tomasin19
fan-fricken-tastic such a great well planned guide
Glad you liked it, thank you!!!
this could have application far beyond just 3d rendering. Conversion of the image into 3d object is effectively object detection algorithm for self driving vehicle. I could see stable diffusion leak into self driving vehicle sector in the future.
that's thinking so far ahead my car crashed twice just thinking bout it.
Sounds super interesting tho, even if I have absolutely no idea how hahahhahha
@@Not4Talent_AICurrently in the self driving tech they mainly use object detection from the camera frame.
Then attempt to project detection coordinate onto a xyz coordinate frame. Then the car operate from xyz coordinate frame. It has similar issue of temporal and spatial inconsistency that AI video have. IE: if a car went behind truck, the algorithm doesnt know it existed anymore. There is a lot of back end code that try to stitch together detection across multiple frame to maintain object temporal consistency.
This workflow of image to 3d blender effectively convert image or video data directly into 3d object. So we don't have to stitch together data to provide temporal consistency.
A lot algorithm is share between the AI 3D generation space and object mapping space. Since gen AI is converting from coordinate (or lower dimension promt) to 3d object, and 3d object detection is the opposite workflow, converting from picture to 3d space.
oh wow that's extremely insightful and cool ngl. Wasnt expecting to learn about this today hahahaha ty for the detailed explanation!!
I'm not gonna say if you are right or not cuz I have 0fking idea, but you seem very very knowledgeable on the topic. So amma trust you hahaha
Would love to see if this goes somewhere ngl@@keenheat3335
@@keenheat3335 nice input htere. i knew convolution neural network was used as part of image identification process so it make sense as it is at the root of the caption process of images in SD. i don't dwelve into that but i see te point where a 3D depth combined is much more recognizable than a flat 2D shape.
good luck on your work, i could have just been a bad idea from the start but history have shown that discovery are made by accidents sometimes.
The topic of consistency in AI is very interesting! Keep going, another amazing video💪🏻
thank you!!! hope it helps
Excellent video my dude, gonna have to bookmark this one for later, because I'll have a use for this knowledge in the future.
Any checkpoints (anime and/or realistic) in particular you'd recommend for environments?
thank you so much!!
I usually just use bout 3 or 4 models:
realistic 1.5: EpicRealism
realistic SDXL: ColossusProject xl
anime 1.5: Anylora or ivory
AnimeSDXL: (haven really used this much, probably better ones are out there) animagine xl
@@Not4Talent_AI Cheers mate
sorry i'm late to the party, i had to motivate myself.
that's a lot of decation for the niche audience that would be willing to use it.
but now thanks to you i managed to understand what i missed from active motion picture instruction on depth map inside blender ! and there is much more than what i was expecting.
i'll be waiting your next video. i would definitly need that. My 3D waifu never managed to drink the beer in my Lora Samples !
thanks!!
Idk if my next vid will touch much 3D, didnt find it necesary to use it as a way to integrate characters.
Maybe i'll teach my version of clip drop relight at home, but I'll see.
I still need more blender practice to make tutorials on it hahahahaha. What was it that you where expecting to use it for btw? Just in case I can add it or I can make a separate short/video on it
@@Not4Talent_AI the ai feels you that we see often ? It remain.
Making it appear dimer with a custom character/style is a nice way to prevent part of it. However the next step is to control the environnement, the object interaction and the lightning of each scene and make it fit acurately. (Your use of reference/IP adapter was a great point toward that).
Changing character shape and shade with the coherent background is Key to make a great masterpiece. One that i would put as desktop wallpaper without anyone noticing the ai things. (We are far away still :) )
Yass! for now I'll focus on adding the character, and trying my best to make it integrated into the scene, with proper lighting, prespective, scale, color... etc
As well as the character being able to sit, grab, break... parts of it. @@lefourbe5596
@@Not4Talent_AI as Always thanks you for your awesome dedication. I Hope you find much joy in the process.
ty for helping and watching!! And yeah, exploring this stuff is super fun@@lefourbe5596
Great tutorial! Your clear instructions and simple approach make tricky concepts easy to grasp. Excited to test these techniques in my projects. Keep it up! 👍
Thank you so much! Hope they help 😄
Wow, your video came at the right time. I've been trying to find this method all day🤩
nice!! glad to hear that hahahaha I hope it works well for you!
Holy shit, feels like that video cam after my feedback that I can't apply that knowledge to the game development ^^
Thanks for showing how to create my own HDR textures with SD!! That helped a lot! If only I'd stop struggling with the basics of Lora and In-/Outpainting, then u would be able to apply this to my job.
And also thanks for showing us the 3D modelling AI! I hope there may exist some locally install-able AIs in the future, just like SD 🙏
And I appreciate so much that you give us free access to your videos! They are definitely high quality and contain a lot of effort. I wish I could donor smt. in return, but as a student in game developmeny my pockets are kinda empty ..
happy to hear it helps!!! If you have any issues with inpainting or outpainting, just ask! After posting a video I usually help around in the discord if I can, so it's a good time to ask any questions hahaha
Also, no need to donate at all, comments like this already make it all worth it. As long as the vids I put out help yall, thats more than enough!
as always, thank you so much for your kind words!
@@Not4Talent_AI This mindset is rare. You are such a good fellow. Thank you!
You really don't mind if I ask you stuff via DM? (I mean I can trade it for Game Dev and Coding knowledge ;p)
hahahaha ofc I dont mind! @@mysticalfoxie
creative thinking on blender and stable diffusion combination
Tyy!!!
Thanks for the mention!
BTW, Euler - pronounced Oiler :)
Also, I haven't even thought about using this for unique Skydomes for games :o should be interesting
and also,
"But sir, I want to type words and get a full 3D environment in seconds 😡"
hahaahahhahha thats for next time XD
and ty for the pronunciation tip, I'll try to keep it in mind next time too hahhhahah
YEAH me too just make it fast bruh X).
if only... *pepe sad*
I want to subscribe to your channel again. 😮
hahahahah tyty :3
🔥🔥🔥🔥🔥
Excellent....
tyty!
You always come up with interesting workflows. Thanks man. 👍👍
Tyty!
very cool, brilliant summary once again
Tyty!!😁
You are my idol
Hahahahha ty do much!!!
Amazing
Thanks!!
What is the maximum resolution of a 360 image with a 4090? Thanks bro
Directly from text to image or with image to image without extra upscalers?
In txt2img you could do lile 1024x2048 and then highres to 1.6 or so I think.
In img to image similar sizes.
If you want you can upscale with ultimate sd upscale or rezise with tiled diffusion. There you pretty much go 8k if you want
oh yea we slopmaxxing
XDD
08:52 what secret formula :)?
idk, ask him. If I knew I'd use it too hahahhaha
(its most likely not even secret, he just has different settings than mine but I didnt ask for them xD)
2:23
now I'm worried, I dont see anything weird XD aside from the face being al fked up cuz I didnt upscale the image hahahaha
@Not4Talent_AI I don't see anything wrong as well ^^
He probably just tried to scare you 😂
How are you, a 3D modeler, so hyped for this? Hyped for something that can replace you?
I don't think it can. As a 3D modeler I have the skills to adapt to it and implement it as a part of my workflow. I'm hyped for it to save me time in the mundane aspects of the job, and hyped to learn a new tool that has never been explored before.
If at some time it is good enough to replace me entirely, I have other skills to keep the job. so I'm confident and lucky enough to not have to worry. Ofc if my situation was different, I might not be as hyped.
@@Not4Talent_AI Great answer, keep up the good work 👌
btw do you have a portfolio or something, where I can checkout your 3D work?
thanks!!
Tbh, not really hahhaahha the only portfolio I have and can show is probably like 4 years old... so I'd rather not share XD
And the rest of my asset-focused portfolio is based on projects I've worked on and I'm unsure if I can legally show them.
I have a 3D printed hollowknight figurine that I made which I could take a pic of I guess XD
I mainly do lowpoly assets for games and web tho @@notDreadful
@@Not4Talent_AI That's alright 😁👍
@@Not4Talent_AI You should give yourself more credit, besides your 3D portfolio is bound to be better than mine, which just consists of terrible 3d animations from like 2009.