正在加载视频...
视频加载失败
Dude GPT-Image-2... wtf... how.
32 条评论
😉
It's not that crazy... they steal/train on the actual images that include barcodes lol
oh thats insane. They mentioned QR codes but I didn't think of barcodes. Wonder what other kinds of things like this it can do!?
I really think that tomorrow chatgpt 5.5 will absolutely cook claude 4.7
it did this without me asking:
it’s the most freakish geeky crazy attention to details, total OMG 😳 and bravo to OpenAI. And for you noticing
Using it feels like a Superpower. I never thought image models could produce such great results under a variety of constraints
what the fuck
Actually wild!! Like this means it able to truly (kinda) represent its training data.
Why is that crazy to you? They're preprocessing your request with an LLM, which researches the correct ISBN, probably calls a barcode generator tool and THEN lets the image generator put that into the image.
I dropped this photo (225x225px) and said to make it look better. Ok it messed up the numers but adding text that wasn't there, was surprising. Not my hand btw 😅
i bet it can't do qr codes
least impressive thing i've seen. UPC codes can be generated by a label maker for gods sake
Things are really... intense in a good way!!
Woah - that's pretty incredible. How can it preserve scan-ready structure so well? What does this say about compression? I guess a very large model can encode an enormous amount of recurring structure in a compact statistical form.
World model confirmed?
I think OAI has the best imagen model even though I don't like the company.
maybe they're providing it with tools to generate different types of codes.
Yo that’s crazy! ChatGPT has Claude beat in terms of picture for sure.
at this point im gonna start asking for a image with the fixes for my codebase
How on earth did you decide to give that prompt 😅
works for QR codes too!
What do you mean, it is trained on that... you gave it the easiest prompts of it's life
crazy af
that’s wild
Don't call it a comeback!
Wow the computer did what you asked it to. Pretty much AGI
How
Wild time to live in
The amazement people get out of basic shit done by an AI model is amazing in it self
omg
barcodes are way less complicated than letters
![Tutorial: how to make equirectangular panorama with GPT Image 2 + Codex Step 1: generate an image with GPT Image 2 prompt: make equirectangular panorama of [PLACE] Step 2: feed your panorama to Codex as a reference and prompt: make mouse-controlled 3D view I'd love to see your experiments!](https://image.24vids.com/tw-2046925601773113689/amplify_video_thumb/2046923757747363840/img/PZR2c54L2p-Jtssp.jpg)
![GPT Image 2 is insanely good...I generated a 360° equirectangular panorama in Happycapy with just a skill + prompt. Step 1: Select the generate-image skill Step 2: Enter a prompt like: “Use a frontend 360 viewer to display an equirectangular image of […] using the GPT-Image-2 model.” Wanna see how you all get creative with this](https://image.24vids.com/tw-2047214663721681288/amplify_video_thumb/2047213896185110528/img/Bwng_H72deHsJgqK.jpg)
