Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

ChatGPT Images 2.0

openai.com

Post reply on HN
Print view View original

481–490 of 1001 posts

Re: ChatGPT Images 2.0

#481

Post by VladVladikoff »

vunderba wrote: post #83

OpenAI’s gpt-image-1.5 and Google’s NB2 have been pretty much neck and neck on my comparison site which focuses heavily on prompt adherence, with both hovering around a 70% success rate on the prompts for generative and editing capabilities. With the caveat being that Gemini has always had the edge in terms of visual fidelity. That being said, gpt-image-1.5 was a big leap in visual quality for OpenAI and eliminated m…

Why does Gemini 3.1 get a pass for the same reasons they got image 2 gets a fail on the flat earth one? Gemini has all sorts of random body parts and limbs etc.

Re: ChatGPT Images 2.0

#482

Post by jdironman »

simonw wrote: post #37
simonw wrote: post #15

I've been trying out the new model like this: OPENAI_API_KEY="$(llm keys get openai)" \ uv run https://tools.simonwillison.net/python/openai_image.py \ -m gpt-image-2 \ "Do a where's Waldo style image but it's where is the raccoon holding a ham radio" Code here: https://github.com/simonw/tools/blob/main/python/openai_imag... Here's what I got from that prompt. I do not think it included a raccoon holding a ham radio…

I just got a much better version using this command instead, which uses the maximum image size according to https://github.com/openai/openai-cookbook/blob/main/examples... OPENAI_API_KEY="$(llm keys get openai)" \ uv run 'https://raw.githubusercontent.com/simonw/tools/refs/heads/main/python/openai_image.py' \ -m gpt-image-2 \ "Do a where's Waldo style image but it's where is the raccoon holding a ham radio" \ --quali…

The real NFTs where the images we generated along the way

Re: ChatGPT Images 2.0

#483

Post by ComputerGuru »

ea016 wrote: post #13

Price comparison: GPT Image 2 Low : 1024×1024 $0.006 | 1024×1536 $0.005 | 1536×1024 $0.005 Medium : 1024×1024 $0.053 | 1024×1536 $0.041 | 1536×1024 $0.041 High : 1024×1024 $0.211 | 1024×1536 $0.165 | 1536×1024 $0.165 GPT Image 1 Low : 1024×1024 $0.011 | 1024×1536 $0.016 | 1536×1024 $0.016 Medium : 1024×1024 $0.042 | 1024×1536 $0.063 | 1536×1024 $0.063 High : 1024×1024 $0.167 | 1024×1536 $0.25 | 1536×1024 $0.25

It can generate 3840x2160

Re: ChatGPT Images 2.0

#484

Post by henry-j »

debazel wrote: post #462
madrox wrote: post #202

This seems like a great time to mention C2PA, a specification for positively affirming image sources. OpenAI participates in this, and if I load an image I had AI generate in a C2PA Viewer it shows ChatGPT as the source. Bad actors can strip sources out so it's a normal image (that's why it's positive affirmation), but eventually we should start flagging images with no source attribution as dangerous the way we flag…

> but eventually we should start flagging images with no source attribution as dangerous the way we flag non-https. Yes, lets make all images proprietary and locked behind big tech signatures. No more open source image editors or open hardware.

C2PA is actually an open protocol, à la SMTP. the whole spec is at https://spec.c2pa.org/, available for anyone to implement.

Re: ChatGPT Images 2.0

#485

Post by vunderba »

VladVladikoff wrote: post #481
vunderba wrote: post #83

OpenAI’s gpt-image-1.5 and Google’s NB2 have been pretty much neck and neck on my comparison site which focuses heavily on prompt adherence, with both hovering around a 70% success rate on the prompts for generative and editing capabilities. With the caveat being that Gemini has always had the edge in terms of visual fidelity. That being said, gpt-image-1.5 was a big leap in visual quality for OpenAI and eliminated m…

Why does Gemini 3.1 get a pass for the same reasons they got image 2 gets a fail on the flat earth one? Gemini has all sorts of random body parts and limbs etc.

That's a mistake~ None of the models successfully passed the Flat Earth composition test. I've updated the passing criteria to be more explicit as well. Thanks for catching that!

Re: ChatGPT Images 2.0

#486

Post by what »

vunderba wrote: post #83

OpenAI’s gpt-image-1.5 and Google’s NB2 have been pretty much neck and neck on my comparison site which focuses heavily on prompt adherence, with both hovering around a 70% success rate on the prompts for generative and editing capabilities. With the caveat being that Gemini has always had the edge in terms of visual fidelity. That being said, gpt-image-1.5 was a big leap in visual quality for OpenAI and eliminated m…

Where can I see the actual prompts and follow ups you fed each model?

Re: ChatGPT Images 2.0

#487

Post by henry-j »

mdasen wrote: post #337
madrox wrote: post #202

This seems like a great time to mention C2PA, a specification for positively affirming image sources. OpenAI participates in this, and if I load an image I had AI generate in a C2PA Viewer it shows ChatGPT as the source. Bad actors can strip sources out so it's a normal image (that's why it's positive affirmation), but eventually we should start flagging images with no source attribution as dangerous the way we flag…

> Bad actors can strip sources out I think the issue is that it's not just bad actors. It's every social platform that strips out metadata. If I post an image on Instagram, Facebook, or anywhere else, they're going to strip the metadata for my privacy. Sometimes the exif data has geo coordinates. Other times it's less private data like the file name, file create/access/modification times, and the kind of device it wa…

There’s actually a part of the NY state budget right now (TEDE part X, for my law nerds) that’d require social media companies to preserve non-PII provenance metadata and surface it to the user, if the uploaded image has it.

linkedin already does this--- see https://www.linkedin.com/help/linkedin/answer/a6282984, and X’s “made with ai” feature preserves the metadata but doesn’t fully surface it (https://www.theverge.com/ai-artificial-intelligence/882974/x...)

Re: ChatGPT Images 2.0

#488

Post by kibibu »

bulletsvshumans wrote: post #241
kibibu wrote: post #107

Genuine question: what positive use cases are sufficient to accept the harm from image generators? One that i can think of: - replacing photography of people who may be unable to consent or for whom it may be traumatic to revisit photographs and suitable models may not be available, e.g. dementia patients, babies, examples of medical conditions. Most other vaguely positive use cases boil down to "look what image gene…

Democratizing visual communication is arguably useful, for instance helping people to create diagrams that illustrate a concept they wish to convey. This is contingent on the tech working sufficiently well that the visuals are more effective at communication than the text that went into producing them though.

I'm not convinced that "arguably useful" is sufficient to offset its much more heavily-used application as a casually-available disinformation engine.

I mean, the cat's out of the bag; but the cat stinks.

Re: ChatGPT Images 2.0

#489

Post by CamperBob2 »

vunderba wrote: post #469
CamperBob2 wrote: post #467

Earlier quoted context omitted.

It'd be interesting if you could add HunyuanImage-3 to the competition. It's better than Z-Image at almost everything I've thrown at it. It can be (slowly) run at home, but needs 96GB RTX 6000-level hardware so it is not very popular.

I’ll have to give it another try. Its predecessor, Hunyuan Image 2.0, scored pretty poorly when I tested it last year: 2 out of 15, so it'll be interesting to see how much it has improved. Here's ZiT, Gpt-Image-2, and Hunyuan Image 2 for reference: https://genai-showdown.specr.net/?models=hy2,g2,zt Note: It won't show up in some of the newer image comparisons (Angelic Forge, Flat Earth, etc) because it's been depreca…

It does quite a bit better than 2.0, I think. Or at least it may be stylistically different enough to justify a rematch against the others.

Ring toss: https://i.imgur.com/Zs6UNKj.png (arguably a pass)

9-pointed star: https://i.imgur.com/SpcSsSv.png (star is well-formed but only has 6 points)

Mermaid: https://i.imgur.com/R6MbMPX.png (fail, and I can't get Imgur to host it for some reason even though it's SFW)

Octopus: https://i.imgur.com/JTVH7xy.png (good try, almost a pass, but socks don't cover the ends of all the tentacles)

Above are one-shot attempts with seed 42.

Re: ChatGPT Images 2.0

#490

Post by breezybottom »

ndriscoll wrote: post #219
mcmcmc wrote: post #159

Earlier quoted context omitted.

So the use case is just IP theft so you can get more Paw Patrol? AI aside, if you’ve truly exhausted all the simple readers, maybe she should move on to more advanced books instead of repeating more of the same and gamifying it, which seems a great way to destroy a child’s natural curiosity.

Sure, I don't view "IP" as valid, don't entertain the idea that it is possible to "steal" it, and absolutely don't care that someone out there might be sad imagining me making a coloring book for my kids. In fact I'd go so far as to say that holding the position that there's something wrong with tailoring teaching to a child's interests and avoiding that for fear of copyright concerns of all things actually makes you…

Ah the old sovereign citizen reverse uno. It's actually evil NOT to use the art theft machine to dumb down your children.
Post reply on HN