2D image to 3D travelling, in less than 5 minutes: Blender, Automatic1111, Stable Diffusion
ฝัง
- เผยแพร่เมื่อ 26 มิ.ย. 2023
- Welcome to this quick tutorial to turn your single image into a great 3D camera travelling with parallax, light, and depth of field.
Automatic1111 extension is stable-diffusion-webui-depthmap-script github.com/thygate/stable-dif...
![How To Create PERFECT Depth Maps In Blender [COMPLETE TUTORIAL]](/img/n.gif)
Camera projection is one of the most powerful VFX trick I have learnt over time.
Easy on hardware with good results 10/10
Thank you, I agree 😁
But you can also mask the image into three layers in Photoshop and arrange the split three images one after another in blender, so that it shows a parallax effect.
Yes, but you would need to split using lasso (good luck, sometimes it's hard) and "invent" the back of 2 layers using the clone tool or else. In the end it's far longer and you won't also get the smooth DoF since you only have 3 clear layers, not a real 3D based on the depth map.
this is the best tutorial. i have yet to learn camera compositing but there are so manytools to still learn. thanks
Thank you! Share with your friends :)
Oh my gosh!Very amazing and free my working time on scene animation
Exactly! IA is a tool. We can use it to push our art forward.
The video is amazing please do more videos on this topic, not just landscape please add people and animals
My friend.. You've made an amazing tutorial. Instant sub with the bell too.
Thank you very much! Feel free to share 😀 Have a great day!
Your videos are wonderful, make more
Thank you very much @sino1449 Feel free to share :)
yeah i am subbing this channel
no one can stop me
not even thanos
btw its pretty funny how that ai voice sounded really realistic
Thanks guys! Feel free to share with your friends :)
This is so cool!!!! & I do love Blender as well, great combo of Automatic111 & Blender. I wonder if there's ways to further refine the mesh with a few more automatic steps? I can do it by hand, but would be nice if it cut out whats behind prominent objects, so we can pan & tilt the camera a bit more. either way its already a great shortcut & can be refined by hand, still having saved tons of time. Thanks for the great tutorial, I had looked at that extension several times & never downloaded, because I didnt know all we could do. Subscribed!
Thank you very much. Refining the mesh in Blender is easy: subdivide modifier. Sure, turning a 2D image into something even more 3D automatically would be great. It means the AI would understand what's in the image (rocks, distant clouds, ...). It's already coming with other AIs isolating 3D objects, AFAIK not a full scene yet with already rigged trees and grass ... yet :)
@@Realistis Thats true, exciting times we're in, every day seems a new breakthru. Eventually im sure a Blender Ai comes, that with our guidance knows how to use its own toolset, write its own scripts, this way it can generate objects & scenes rather than pixels. Im sure it will come eventually, blender is by far the best place for such a thing, being open source & so broadly featured. There is a SD texture addon already, but I mean more a Blender AGI lol😁
@@macronomicus If an AI could write Blender scripts based on another AI trying to get result in it (txt2img2script), that would be great 😁
At some point, since we'll all get the same "art power", we'll get bored with art pretty fast (Yeah-I-can-do-it-too), I wonder what would be distinctive and differentiate an artist from others.
Thank you for the great tutorial.
Do you know how to export this kind of texture from blender to 3dsmax ?
I tried with alembic and obj, but i can't get the texture.
Merci 😉
Nope, sorry. You could try to first map the image on the mesh by unwrapping the UV? It should produce a MTL file to manage to texture.
Avec plaisir 🙂
My mesh has some "holes" in it. Like the edges of the objects are missing. Does someone know how to fix this?
Not sure if I understand it well but, did you UNcheck remove occluded edges? See th-cam.com/video/2aeNG8u6lJg/w-d-xo.html
I followed along and all I get is black, bright pink, or white when I try to add the image. I can get an image on a flat plane. What am I missing?
Usually pink mesh means you have an issue with your shader. Did you start with default Blender settings? Or did you already use a default file?
Let's start with the beginning:
Double check webui settings before generating and exporting.
Then in Blender > Shading workspace, double check: attribute color > emission > Material output (Surface, not volume).
There are so many parameters in Blender, it's hard to know what is different from other setups, but you can check a few things: EEVEE or Cycles render engine (not workbench or else), world shader must be default background > World output (surface), ...
I hope it helps you debug. You can search for pink texture blender
My color attirbute not working for material how can i apply it to 3d mesh ?
Could you tell me a little bit more about your settings? Did you use an emission shader? Double check from th-cam.com/video/2aeNG8u6lJg/w-d-xo.html
how do you do actors voice sound? can you do a tutorial on that? his voice sounds good.
Thanks, but this is IA 😅 I'm better at Blender than speaking English.
@@RealistisSame with speaking english here. But I understood that it is ai. I meant what ai services did you use. I guess it’s an interesting topic on that own and a tutorial could be a good content for your channel as well
@@Bugulab It's an elevenlab free voice with a few tweaks (you really have 2 parameters). Nothing worth making a tutorial I think.
How about using my own painted background....will it work? 🤔
Yes, any image with some depth will be interpreted as a depth map, and you'll be able to get a 3D projection. You could also paint your own depth map and get the AI to mesh it then import it into Blender.
Can you provide me a link to 1111 plug-in please?
Please see in description
I just get orange noise instead of the image
With attribute color?
Looks cool, but it's sad that Stable Diffusion is based on stolen artworks and using it is unethical
like no artist ever steal idea
@@Whatnexttttt No, most artists don't steal.
If you mean inspiration, it's a different, natural thing for every human, and before you write "AI takes references and is inspired like humans!" - No, image generators are soulless algorithms which don't even understand anything, they don't have consciousness nor anything which would resemble a real, thinking and living being
@@Whatnexttttt artists use references or draw from already existing stuff, yes
however no artist stoles other's artists artwork, mesh it a little and call it new
and its even worse when the ai does a perfect replica of the original artists art, you're basically just destroying someones career just because
no problem just use models that are trained on non "stolen" artworks. Stable Diffusion is innocent.
@@johngate70 Stable Diffusion is exactly trained on stolen artworks. Unfortunately there are no models trained on copyright free data