211–220 of 718 posts
This thing's ability to produce entire infographics from a short prompt is really impressive, especially since it can run extra Google searches first. I tried this prompt: Infographic explaining how the Datasette open source project works Here's the result: https://simonwillison.net/2025/Nov/20/nano-banana-pro/#creat...
Earlier quoted context omitted.
Some days it feels like I'm the only hacker left who doesn't want government mandated watermarking in creative tools. Were politicians 20 years ago as overreative they'd have demanded Photoshop leave a trace on anything it edited. The amount of moral panic is off the charts. It's still a computer, and we still shouldn't trust everything we see. The fundamentals haven't changed.
Politicians absolutely were doing this 20-30 years ago. Plenty of folks here are old enough to remember debates on Slashdot around the Communications Decency Act, Child Online Protection Act, Children's Online Privacy Protection Act, Children's Internet Protection Act, et al. https://en.wikipedia.org/wiki/Communications_Decency_Act
I...worked on the detailed Nano Banana prompt engineering analysis for months ( https://news.ycombinator.com/item?id=45917875 )...and...Google just...Google released a new version. Nano Banana Pro should work with my gemimg package ( https://github.com/minimaxir/gemimg ) without pushing a new version by passing: g = GemImg(model="gemini-3-pro-image-preview") I'll add the new output resolutions and other features ASAP…
https://minimaxir.com/2025/11/nano-banana-prompts/#hello-nan...
My recreations of those pancake batter skulls using Nano Banana Pro: https://simonwillison.net/2025/Nov/20/nano-banana-pro/#tryin...
The interesting tidbit here is SynthID. While a good first step, it doesn't solve the problem of AI generated content NOT having any kind of watermark. So we can prove that something WITH the ID is AI generated but we can't prove that something without one ISN'T AI generated. Like it would be nice if all photo and video generated by the big players would have some kind of standardized identifier on them - but now you…
It solves some problems! For example, if you want to run a camgirl website based on AI models and want to also prove that you're not exploiting real people
So, you exploit real people, but run your images through a realtime AI video transformation model doing either a close-to-noop transformation or something like changing the background so that it can't be used to identify the actual location if people do figure out you are exploiting real people, and then you have your real exploitation watermarked as AI fakery.
I don't think this is solving a problem, unless you mean a problem for the would-be exploiter.
Earlier quoted context omitted.
> It's still a computer, and we still shouldn't trust everything we see. The fundamentals haven't changed. I think that by now it should be crystal clear to everyone that it matters a lot the sheer scale a new technology permits for $nefarious_intent. Knives (under a certain size) are not regulated. Guns are regulated in most countries. Atomic bombs are definitely regulated. They can all kill people if used badly, th…
I think we're overreacting. Digital fakes will proliferate, and we'll freak out bc it's new. But after a certain amount of time, we'll just get used to it and realize that the world goes on, and whatever major adverse effects actually aren't that difficult to deal with. Which is not the case with nuclear proliferation or things like that. The story of human history is newer generations freaking about progress and nov…
We could use the opportunity to deploy robust systems of verification and validation to all digital works. One that allows for proving authenticity while respecting privacy if desired. For example… it’s insane in the US we revolve around a paper social security number that we know damn well isn’t unique. Or that it’s a massive pain in the ass for most people to even check the hash of a download.
Guess which we’ll do!
The interesting tidbit here is SynthID. While a good first step, it doesn't solve the problem of AI generated content NOT having any kind of watermark. So we can prove that something WITH the ID is AI generated but we can't prove that something without one ISN'T AI generated. Like it would be nice if all photo and video generated by the big players would have some kind of standardized identifier on them - but now you…
This watermarking ceremony is useless. We will always have local models. Eventually the Chinese will release a Nano Banana equivalent as open source.
If watermarking becomes a legal mandate, it will inevitably include a prohibition on distributing (and using and maybe even possessing, but the distribution ban is the thing that will have the most impact, since it is the part that is most policable, and most people aren't going to be training their own models, except, of course, the most motivated bad actors) open models that do not include watermarking as a baked-in model feature. So, for most users, it'll be much less accessible (and, at the same time, it won't solve the problem.)
“Generate an image of an african elephant painted in the New England flag, doing a backflip in front of the russian federal assembly.”
OpenAI made the biggest step change towards compositionality in image generation when they started directly generating image tokens for decoders from foundation llms, and it worked very well (openais images were better in this regard than nano banana 1, but struggled with some OOD images like elephants doing backflips), but banana 2 nails this stuff in a way I haven't seen anywhere else
if video follows the same trends as images in terms of prompt adherence, that will be very valuable... and interesting
Earlier quoted context omitted.
And that's not a good thing.
It depends on how you're looking at it. For the people not getting handed counterfeit currency, it's probably a good thing.
SynthID seems interesting but in classic Google fashion, I haven't a clue on how to use it and the only button that exists is join a waitlist. Apparently it's been out since 2023? Also, does SynthID work only within gemini ecosystem? If so, is this the beginning of a slew of these products with no one standard way? i.e "Have you run that image through tool1, tool2, tool3, and tool4 before deciding this image is legit…
No, its not the beginning, multiple different watermarking standards, watermark checking systems, and, of course, published countermeasures of various effectiveness for most of them, have been around for a while.
I...worked on the detailed Nano Banana prompt engineering analysis for months ( https://news.ycombinator.com/item?id=45917875 )...and...Google just...Google released a new version. Nano Banana Pro should work with my gemimg package ( https://github.com/minimaxir/gemimg ) without pushing a new version by passing: g = GemImg(model="gemini-3-pro-image-preview") I'll add the new output resolutions and other features ASAP…
Minor clarification, the cost for every input image is $0.0011, not $0.06.