Text to 3D is AWESOME now! - AI Tools you need to know
Vložit
- čas přidán 15. 05. 2024
- Text to 3D is awesome now. Check out these AI Tools you need to know. The can generate fully texutred and rigged 3D models from Text, Images, Photo or Sketch. The future of AI is bright and beautiful.
#### Links from my Video ####
StableProjectorz • StableProjectorz v 1.2...
stableprojectorz.com/
www.meshy.ai/
3d.csm.ai/
gala3d.github.io/
lumalabs.ai/
binary-opacity-grid.github.io/
zhenglinzhou.github.io/HeadSt...
#### Join and Support me ####
Buy me a Coffee: www.buymeacoffee.com/oliviotu...
Join my Facebook Group: / theairevolution
Joint my Discord Group: / discord
AI Newsletter: oliviotutorials.podia.com/new...
Support me on Patreon: / sarikas
00:00 Intro
00:45 LumaLabs Text to 3D
03:42 Meshy Text to 3D
05:42 CSM 3D Ai
06:56 Photo to 3D
07:32 HeadStudio 3D Face Render
08:23 StableProjectorz - Jak na to + styl
As an aspiring later in life 3D artist with 9 years of Blender tutorials and personal projects under his belt, it's nice to yet again see my ambitions be crushed like a bug.
... same here. Creativity is not longer a skill, it just cost credits
Creativity still matters at least for now. However, creatives should stay on top of the latest AI advances and use them to their fullest.
I'm in the same boat. At the moment, most of what I'm seeing wouldn't be high enough quality to put in my portfolio even though I'm more at an indie level than AAA. However, it's getting there. My biggest fear is that I won't be able to use these tools myself, either because of regulations or cost. If it gets to the point where I can quickly and easily create my own AAA custom movies and video games by myself, that's not a world I necessarily hate, even if I can't get paid for those skills anymore.
It's not about AI taking over what you will do as a 3D artist. Learn how to make AI an asset in the type of work you will be doing.
"It's not AI that takes over, it's humans that leverages AI who will"...
Imagine if you’ve been doing 3D/CG for more than 30 years…like myself.
As an Actual 3D artist these are looking great for crowd characters, or things in a secondary plane. For an actual character that will be at first plane will be garbage, maybe in some months or next year will generate things that can be used in Industry.
for solo indie its miles useful for those one off games
@@user-qr4jf4tv2x yeah but at the end will be double work, because will need to replace all those characters for better ones. To early stages of development are useful to test in engine stuf.
I see this being useful for quickly making rough starting shapes from a sketch allowing you to jump into fine detail faster. Just imagine how much clicking and dragging is saved.@@angeldart3D
Most work in the future will be creating crude products with AI, and refining them by hand instead of doing the whole thing from the ground up
@@rodrigoibaceta3938 this is the work now, in the future (not "too" future) ai will just do all the stuff and we'll be (almost) all unemployed
I clicked because the title said this is clickbait. I appreciate the honesty.
I put that in as a test out of three thumbnails to if it works 😅👍
@@OlivioSarikasyes a scam
Which version got the most clicks? @@OlivioSarikas
You've alerted me to some fantastic tools, thank you! Other videos didn't have a lot of these. Subbed.
I worked with 3D AI STUDIO which also does this and was able to 3D print my model.
Meshy worked also great!
now you wont work anywhere since they dont need you anymore, but hey you got your model 3d printed right
I’ve been texturing VRChat avatars with stable diffusion for well over a year. I love where this stuff is going! I hope more tools get AI assistance cuz Unity stuff hurts my poor smooth brain
Wow so many clueless fools who think AI is helping them. AI does not create, it takes and generates! It's killing creativity.
@@lynkylo5530 womp womp
@@koopakid2002 womp womp somebody isn't an artist
I tried Luma last week, just the basic version and it was okay but you clearly needed to pay to get anything of use or quality out of it. Which is fine, I just wasn't ready to pay yet as I don't have a current use for it besides messing around. In the future it will be an option and it has a lot of potential.
Very cool! Thanks for sharing.
Can it do the rear perspective of an image?
Super usefull! Thanks for the overview!
Thanks! Useful information. 👍👍👍
Thank you
My go-to channel for AI news, you're always so damn fast with new the newest developments in AI!
Thank you very much :)
Fantastic work!
Love your videos my man
Had a super quick question
Can 3D image generation be achieved locally?
Really appreciate you getting back. Thanks again
Simple rule: if model is Open Source than you can run it locally ;-)
Looks sooo fuuun! Thx for the informative upload..
Hello teacher, I would like to ask you, I used Photoshop to splice the head and body, but the skin color of the person is different now! How can I unify the skin color of the person in stabledifdusion, and still retain the original light and shadow of the person! I hope the teacher can give me an answer, thank you very much!
4th dimension? I'm sold! 😅
Are these models water tight?
Thank you!!! 🙂
Grate video 👍 love how AI is making all this cool improvements in graphics and 3d 👍🙂
stable projector looks awesome
...this week! The texturing tool is really cool.
Very interesting run down. Next time, I hope that you will be able to output and STL and show us a sample print.
thank you for this video!
I want to use this for my 3D printer.
If I had an AI that knows what it's supposed to look like, it can make adjustments on the fly.
In a few years, you'll have it, I bet!
Its already much better than 1 year ago. When they just figure out adding the high detail too, it will be great.
It looks ok on the surface but creates terrible topology that's tough to work with practically in a 3d environment if you want to rig or re-skin etc.
Inaccurate title. Text to 3D is "almost usable" now. Thanks for sharing Stable Projectorz though. That app is the most usable solution around. 😊
Is there any for car parts 3d optimization and generating
GameDev or animatin in the near future: "AI make idea, AI write story, AI make assets, AI make voice and sounds, AI make animations and write code, AI publish new product every hour".
Another great video, thanks Olivio
Thank you :)
Thanks for making more content about AI showing everyone how great it is without even mentionning it's downsides and unethical generation process
Nice video, thanks! Yeah, these tools will help us to realize our ideas faster.
super cool stuff!
Can't wait for robust real open local competition.
Me too
With a requirement of over 40gb of VRAM
When this is ready for local install then I'll be excited:)
Dream Gaussian with mv-dream and imgdream is faster than any of the software's listed on here other than the texturing solution and is about the same quality on github.
Looks decent, but not ready for prime time yet.
Until they can generate high quality 3d models, with correct topology and weight painting (for objects that will be animated), making them ready for game engines with no additional work (which in some cases takes more than just making the model from the start), 3d generating models will not be good enough.
edit: Hopefully 1 more year for this level of 3d generation.
I was waiting for a good AI tool that can create 3D models. This one seems decent. Hopefully, it will get even better.
I'm a solo gamedev. I can code pretty well and do entry level hard surface modelling. So I'm really interested in not paying for all my models because I need consistent styling and I can't get that from marketplaces. But every few months I give these things a shot and find the results are like using raw photogrammetry. Holes, blobs, weird textures. I think that yeah I *could* fix them, but the fixing more or less needs the same skillset as building from scratch and the outcomes are not even close to perfect.
I have also tried getting AI image to generate cool stuff. Feed that image in to a image>>3d modeler and boom... manure. One even gave me a 3d model of a plane with a messed up version of my picture and my picture was just a 3d render.
Learn to retopologize. Its easy
@@neopabo if you're going to learn retopo then you might as well learn how to model in the first place. In fact, you'll never be effective at retopo if you don't understand the fundamentals. So just learn to model.
This tool is so useful for fast prototyping. I repeat. Use it for fast prototyping. Then take it to Blender to recreate your 3D puppet. You are going to save so much time in the concept composition stage. Add these tools to your creative workflow. See the opportunity and the benefit!
This technology has great implications to VR. Creating VR environments to walk around in
is this an end of 3d and vfx ?
Nice video. Well done
you showed very nice programs
i think it holds cool opportunity for everyone who need to complement the skills he isnt used to but also seen some art in the video whos training data might be not 100% legal ... funko and doctor strange e.g. are ip of companies and should not be used for professional work, for hobbiesh projects its totaly fine but A.I. at the moment feels more like thieveries :,) and i dont know how to get rid of this state
Are there any open source products that work with Comfy, Auto1111 or Forge?
If generative AI could be persuaded into producing a series of images of an object, photogrammetry could then make a 3d mesh from the series of images.
Exporting the mesh to STL would easy and/or peazy.
I've played around with COLMAP..it tolerates a fair amount of input image 'noise' while producing decent meshes.
Nothing worth publishing yet, still having fun learning ComfyUI and workflows.
Oh hey thats my crocodile at 3:14 lol
Unless it’s open source local install, I’m not interested.
It will take some time. But i'm hopeful
AI stuff will never be local-install at the time of production, unless you have a mad rig at home, maybe 5-10 years down the line when the hardware has adapted.
Simple Photogrammetry is there, so you can make real photos into a 3d point cloud and then into a mesh, it's a messy process but it works. However an image-to-3d procedural generation on a local hardware is far off.
@@Drak_Thedp I used RenderNet to make images of a Navvy with shovel. I was curious and took that image and used Ollama on my Pi5 to run llava and asked "what is in image ....". It made three paragraphs that were spot on.
@@Drak_Thedp
Wh that's alright. I have patience
I think for games and performance (unless nvidia releases a 64GB user level card) all those scrap textures are going to make devving with this tech hard lol
with in 5 years we will have LLM running locally on our phone
can't wait
Awesome 🔥 Thank^3 to U 👍
Yeah, using that AI for 30 printing, when I work most of the details that you see comes from those textra maps, we take the text from my office off. They're not gonna be half as detailed. So the technology still has a long way to go, so we can take something and put that on the s. DL and 3D printed at home.
I would love to see them create an app for the Apple Vision Pro.
I wonder if these could be directly downloaded into the vision pro and put into the landscape
@@OlivioSarikas There is a program called JigSpace and it says "JigSpace supports all major CAD file types, including OBJ, FBX, STEP, and STL. If your CAD files contain colors, they'll automatically be imported"
Can't wait for Meta to buy them and incorporate it into their VR.
Wow! My two hobbies are meshing 3d printing and AI art.
the generating high resolution option doesn't seem to work. acts like it's doing it but nothing happens. let it go for a long time, never did anything.
Would be nice to drop some 3d examples and based on that get new AI results
The fourth dimension is time, unless you specifically specify spatial 4th dimension. Just saying... awesome video regardless, thank you ;)
in this case i refer to a metaphorical dimesion as both real world and graphics are 3D, this brings it into 3D + Real Life (or simulated AR)
Nvidia Nerf is free to create 3d enviroments. Dome of these have built on that i think
Now give it to me, but on my local PC no web service, like A1111 or ComfyUI but 3D. I have already been experimenting with stable?zero123 but I want this full 3D thing local. Stable projection is nuts...
now we just need molecular transporters and force fields and we can have holodecks
And all we know, this is just the beginning...
One day ai will be so advanced it will recreate new plots in movies on the fly at viewers request and also be able to rerender old games into 3d or certail style and even generate new or endless levels in all directions infinitely based on the style user already played and switch at random like seed or at will or based on all levels predefined.
We are still a few years away from any of these to be usable.
Generate something with them, delete the textures and look at the topology.
Yes, nothing super professional yet. But for the first experiments and little use cases experiments it's already good enough
@@OlivioSarikas not true one of them is but its not in this video ,tripo ai
@@OlivioSarikas Not even a professional stage... If you have any idea how games work the topology is not even good for low end. So many useless quads that would render terrible in game engines. The textures look terrible and out of alignment.
@@jonathanperreault4503 Even Tripo looks bad...
@@helix8847ive been playing with it for a full week and i can tell you its all about the distance of the subject to the camera , if you have a subject that takes a really large part of the image , usually in above the hips portraits then you can get really really nice result ready for 3d printing , on top of that im trying to figure out a displacement method since the texturing is so good
This is amazing for filling my 3D interiors with stuff, for example I can focus on designing the furniture for a kid's room and fill it with AI generated toys, clothes and books. This will save me tons of money on 3D assets.
There goes my career as 3D asset creator, will need to stock up on 3D silicone 1:1 scale figure printing.
@@babbagebrassworks4278 Sounds like a you problem if you're not able to adapt your skill and knowledge to ride whatever wave the future is bringing to us. Think. There's gold waiting for you.
@@TheCynicalCynic Nah, I'm 64 and now have my Raspberry Pi5 with Ollama running 12 LLMs. Next is to get Stable Diffusion running locally. F.. those cloud subscription models. I need 3D digital art for my new 3D resin printer. Got a 3018 laser n CNC mill modded to 30x40. Sell 3D miniatures in a niche market I have 50+ years of knowledge in. I should be fine. If not, I might have to use my 50+ years in robotics and electronics to repair or change the batteries in our mobile overlords. Forget going for the gold, be the one selling the tools. My comment was sarcasm, however life like robotics could be a things as this AI tech crosses into mobile robot capabilities.
If it's any consolation, audiences still generally prefer creations made by humans. CZcams scripts for example... I've generated a number of scripts for CZcams shorts with GPT-4 (via Bing), but the best ones, by FAR (and the ones that have gotten me the most views), are the ones that I've personally written.
You can create great things with AI, but it's hard to gather an audience without that human connection / interaction.
I've had better results with Meshy, especially in terms of texture quality. Aside from the somewhat blobby models, I hope they work on the baked lighting. The textures can look nice but all that baked lighting makes them not great for a variety of lighting conditions.
Good point. Any special prompt or negative prompt words you use to get better results?
For Meshy, I just use a handful of the standard quality tags (best quality, masterpiece, 4k etc.) and then I use the negative prompts they recommend in their prompting docs. Results are still highly dependent on how simple and commonplace the item is (Meshy seems really good with clothing) but the better textures with Meshy give you something you can use for manual touch ups with the clone tool.
As for the baked lighting, you could try some of the lighting tags like flat or diffuse lighting. I'm still new to the tool myself so I don't want to seem like an expert.
Now if only they could come with some better UVs.
Take a sip of your drink, every time he says cool
Now, now, we don't want to create alcoholics here 😅
GREAT😊
It if generate locally it would be good.
I feel like copyright laws are gonna be at war even more so, knowing 3d printing AI art
I think they are using 3d Gaussian splatting from a 2d AI image in the first one.
Ai CAD libraries could be a big deal
AI is the future. All 3D designers will work way faster. It's funny to see how kids on Twitter who are drawing furries are afraid of the technology while every few years many softwares get more advanced to create better things. Nobody will replace professional designers because AI can't create the best models or images if an user has 0 knowledge what makes that render actually good.
Can i tell you something? If we align the world that the ai generates ie how math reps physics through what it is, and how a neuron reps the difference while maintaining what it is, then we should be able to align not only the ai but the ai engines so that the engine is matching the real physics but in its difference so similar to the neuron and the math. If we can integrate the systems to match our current then we could get more accurate predictions. What i mean is the same thing as a scientist learning the differences of the world with his senses, and then the neurons rep the difference that was integrated while being themselves, so a superposition, and then the scientist reflects the neurons through the words and hands to reveal what the mind has revealed. Since the mind has the capacity to align with our current then all is able to be seen or revealed, not immediately like magic but over the difference like expansion through the details, its the sharing of energies that integrates and aligns to reveal the difference. I hope its understandable. What im saying is that we could bring out something far from what we see as real because its not that far, we just are at a different position so the point of view is not aligned. Similar to how humans share and integrate but also can not be aligned and the energies expand until integration of a sort or entanglement.
Little late, buddy this has been around for over a year and a half. And it’s only getting better. Try uploading a drone footage to be blown away and then uploaded to unreal engine.
If you just need generic 3D models, instead of specific bespoke ones that truly support your artistic vision, then you can download models from an online library. You've been able to do this since +- 2001, and the models, if you search for more than a few seconds, are of higher quality than what this "genie" thing generally spits out. So... Talk about solving a non-existent problem with a worse solution. Really on-point tech bro 2024 stuff
Now we need an AI that makes a new platform for games we sell. Since Steam is banning AI games.
Wish these 3d model generators did not require some account that will 100% have its information leaked
wow imagine being born into life, when you reach 5 years old you prompt that you want to become a superhero and youre a superhero, prompt you want to be a policeman and boom ai scenarios at the get go + bionic modifications. Like life is going to get so boring without creativity and problems and everything
7:45 i guarantee they do not have permission from marvel OR benedict cumberbatch to use his likeness like this. remember that ai voice site that got sued into oblivion? this may be next haha
Isn't this old news? I've been using it for several weeks. Have they updated it and made it better?
Mmmmm… the 3d models remind me of CT MRI scans, in the sense the models are a 3d soup of triangles, not nice quads with end loops.
Just saying.
"It's not AI that takes over, it's humans that leverages AI who will"...
I think this and any type of AI that helps people in their respective line of work is exciting indeed.
I'm not sure why you gave any amount of praise to the Meshy tool. Shameful generation quality on both the polygons and the textures.
your use of 4D in the intro is incorrect by every definition of the term.
dali and gala, 2d 3d, husband wife
4:53: downloading a model costs 20 as well. so if you make a model and uprez then download it, it costs 45. a charge for DOWNLOADING - after you've paid $16 per month already? is a bit much to me.
Really? They ask credits for downloads. Yikes. That's not a good look
Why do I even bother? it's more future proof just finding 🏴☠unity/blender assets + plus a bit of work, make something better.
What's worse than subscriptions? CREDITS.
In 10 years craftmanship as we know it today might be approaching it's final days. This is kind of sad
It's not ;-)
Most would be done with macjines but people would still make things (for rich people xD)
looks like we still have to retopologize, sad
Is the model in the thumbnail ai made? I seriously doubt it and if not then that is pure clickbait.
Not one mention of the terrible topology
Yeah no, love your work and have followed your entire course on comfy ui. However you need to introduce some critical perspective, as a 3D artist, these are mostly beyond useless, this is not awesome. The photogrammetry ones are getting there.
Needs knowledge of anatomy
More than that
Bro I have the loras believe me hahahahaha
Yes, lots and lots of female anatomy. For science.
If you are going to use 3D models then you better learn Blender or it will be all for nothing!
I found meshy to be dog doodoo
run because nao gonna to be free forever....
the fact that they charge for those garbage 3d models is puzzling. they should be free i still didnt use them
Well still a long way to call it awesome. :)
AI goober that never moved a vertex in his life.
Seeing the HORRIBLE quality of the 3D models generated by these AIs, one could feel at ease for at least a couple of years, but then you start reading the comments and you remember that people have no kind of artistic criteria. Many nobodies think they're graphic designers just because they have access to AIs that create for them, and you understand that you've wasted 12 years of your life fighting VERY hard to be a professional 3D character modeler, because a bunch of unscrupulous people have come up with the brilliant idea of prostituting 3D creation. Mind you, they haven't focused on making our lives easier by having an AI rig, UV map, or do retopology; no, they've focused on the creative and fun part. Why? Well, I'll leave the answer open for everyone to think about and debate the reason for this.
As for me, after dedicating so much time to this profession, I feel lost because honestly, someone undermining your work in this way is not something I can label as pleasant.