OpenAI’s ChatGPT – 8 New Incredible Features!
❤️ Check out Lambda here and sign up for their GPU Cloud:
Guide for using DeepSeek on Lambda:
Our neural rendering paper:
Sources:
📝 My paper on simulations that look almost like reality is available for free here:
Or this is the orig. Nature Physics link with clickable citations:
🙏 We would like to thank our generous Patreon supporters who make Two Minute Papers possible:
Benji Rabhan, B Shang, Christian Ahlin, Gordon Child, John Le, Juan Benet, Kyle Davis, Loyal Alchemist, Lukas Biewald, Michael Tedder, Owen Skarpness, Richard Sundvall, Steef, Taras Bobrovytsky, Thomas Krcmar, Tybie Fitzhugh, Ueli GallizziIf you wish to appear here or pick up other perks, click here:
My research:
X/Twitter:
Thumbnail design: Felícia Zsolnai-Fehér –
ChatGPT converting images to 3D objects is groundbreaking. Can’t wait to see how Runway’s GEN4 integration boosts creative possibilities. Kudos to the team!
@@CharlotteLopez-n3i not chatgpt
Creativity is dead bud, we just type now 😂
can you 3d print them?
@@pass-the-juice First of all, there are severeral ai models for turning a 2D image none of them have anything to do with openai.
Second, yes you can print them since all of them are gtlf, obj, fbx or stl formated.
Depending on your needs expect some rework to be necessary because none of the 3d generators are flawless and the 2D texture is often used to cover the flaws of the 3D model.
I tried to generate art for a new Balatro joker concept I had but it refused, it explained joker imagery is banned. pretty pathetic imo. also no matter what I tried I couldn’t get it to generate pixel art or 16bit art, the squares were always textured and usually had grid lines. and the colour palette was always way larger than I requested
These AI systems are advancing so quickly, it’s a full-time task just to keep up with what they can do
I feel the same, the FOMO is real with this one.
It is, and currently that’s my task!
And this is so, even though “Moore’s Law” is long dead; imagine if it wasn’t… Is it possible to resurrect it!?
@@akami.7399 same here! I’m thinking how can I use it to create “something” to earn money for a regular income. And while I am thinking that there are surely people already cooking on some serieses, brands, “something”… and tomorrow the internet is fluded with AI generated stuff and It’ll be too late for me.
That is an excellent way to describe it 😂 full time task to keep up, most people around me are very much unaware of what’s happening
the one with drawing blew my mind.
@@oj2416 honestly yeah, it might be a new way to learn drawing in the future!
I’ve been using ChatGPT’s new image generation, and I am really impressed.
Finally, someone who realizes the potential of outpainting and image recontextualizing.
Károly, do you mind explaining the part with that 3D model a bit more? The biggest weakness of the current 3D generators is that they don’t segment the model—it’s only one big mesh with one material. What about the engine you demonstrated briefly? Is it any better? Maybe at least it can use several materials or at least generate more textures?
i like how the cop cars were chasing him backwards
it draws diagrams wrong and draws stuff twice in wrong locations, giving it a selfie makes it copy a person but it often gets the face slightly wrong, as well as fingers. When creating a drawing of laptops it loves to put the screens on the backside for some reason. The diagrams might be better if we iterate on them, and it’s still incredible what you can do with it, especially compared to before. What we now need is a model being able to do this just as good but locally and on one GPU.
what was a bit sad is that it never generates the same image twice. I generated a few iterations on an animefied image, and only the first one looked good. Telling it to use the first result doesn’t work either
I’ve found that it helps to feed it the original image in your next prompt. The results still aren’t perfect, but they’re a lot better.
@RadioInfinity0 i only tried that with the anime prompt but it didn’t really help in that instance
@@kipchickensout Did you try using the Canvas edit feature?
I love the fact that it gets faces wrong, gives me peace of mind actually
Great overview of the new features, this is mindblowing!!
I had recently used ChatGPT to show me what my bathroom would look like remodeled.
I took a picture of it and added inspiration images to the prompt telling it which elements I wanted to incorporate. It was really good.
Modifying it to make it closer to what I wanted took a bit more effort though, and the image processing would often have issues outputting or processing anything. Still, it eventually popped out something that was way better at conveying what I wanted without the work to manually edit it which I would’ve done if this didn’t work.
2:17 That is not “all the intermediate steps”. That is step 1, step 2 broken down into 4 substeps, then a huge jump to like, step 10. It’s still jumping to “the rest of the owl”. I have experimented with this a lot. There is a lack of intermediate steps in the training data. Once you get based a basic block in, it’s very hard to get anything that’s actually the next incremental step and not just a “finished” product
‘Show six steps to drawing an owl’
My favorite image rerun type so far is: made out of tied balloons. What a time to be alive! 🥳🎉🎈
That last thing about the robots is that I think we are very close to this happening. And again you can thanks AI for it. What a time to be alive!
Game developers must be asking themselves whether they should start their projects now or wait for a year. And then a year goes by, and they again ask the same thing.
@@joaodecarvalho7012 lmao same
What a time to be alive!
The changing perspective on the teapot is really impressive. For my scholarly experiment I asked it to make an image of some location, and then told it to zoom in on one one of the buildings in the image. It was able to do it with very little changes to the building, and then easily show the view from the top, front, side etc
This is mind-blowing stuff. So cool to watch!
For some reason I got access to it a little ahead of time and have been playing with it ever since. It does so much more than this. I’m still finding new ways to use it. Love your channel.
What I am more excited about is that in 1-2 years there will be free and open source uncensored alternative to this, which will run on your own computer offline without surveilance