Live data from Hacker News

Nano Banana Pro

blog.google

81–90 of 718 posts

Re: Nano Banana Pro

#81

I've had nano banana pro for a few weeks now, and it's the most impressive AI model I've ever seen The inline verification of images following the prompt is awesome, and you can do some _amazing_ stuff with it. It's probably not as fun anymore though (in the early access program, it doesn't have censoring!)

"Inline verification of images following the prompt is awesome, and you can do some _amazing_ stuff with it." - could you elaborate on this? sounds fascinating but I couldn't grok it via the blog post (like, it this synthid?)

It uses Gemini 3 inline with the reasoning to make sure it followed the instructions before giving you the output image

Re: Nano Banana Pro

#82

I've tried to repaint the exterior of my house. More than 20 times with very detailed prompts. I even tried to optimize it with Claude. No matter what, every time it added one, two or three extra windows to the same wall.

I tried this in AI studio just now with nano banana.

Results: https://imgur.com/a/9II0Aip

The white house was the original (random photo from Google). The prompt was "What paint color would look nice? Paint the house."

Re: Nano Banana Pro

#83

I tried the studio ghibli prompt on a photo my me and my wife in Japan and it was... not good. It looked more like a hand drawn sketch made with colored pencils, but none of the colors were correct. Everything was a weird shade of yellow/brown. This has been an oddly difficult benchmark for Gemini's NB models. Googles images models have always been pretty bad at the studio ghibli prompt, but I'm shocked at how poorly…

Could be they are specifically training against it. There was some controversy about "studio ghibli style". Similarly how in the early days of Stable Diffusion "Greg Rutkowski style" was a very popular prompt to get a specific look. These days modern Stable Diffusion based models like SD 3 or FLUX mostly removed references to specific artists from their datasets.

Re: Nano Banana Pro

#84
post #58

I've had nano banana pro for a few weeks now, and it's the most impressive AI model I've ever seen The inline verification of images following the prompt is awesome, and you can do some _amazing_ stuff with it. It's probably not as fun anymore though (in the early access program, it doesn't have censoring!)

LLMs might be a dead end, but we're going to have amazing images, video, and 3D. To me the AI revolution is making visual media (and music) catch up with the text-based revolution we've had since the dawn of computing. Computers accelerated typing and text almost immediately, but we've had really crude tools for images, video, and 3D despite graphics and image processing algorithms. AI really pushes the envelope here…

I wouldn’t call LLMs a dead end, they’re so useful as-is

Re: Nano Banana Pro

#85
post #61

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

Your use case doesn't even make sense. What customers are clamoring for that feature? I doubt any paying customer in the market for (that product) cares. If the law cares, the law has tools to inquire.

All of this is trivially easy to circumvent ceremony.

Google is doing this to deflect litigation and to preserve their brand in the face of negative press.

They'll do this (1) as long as they're the market leader, (2) as long as there aren't dozens of other similar products - especially ones available as open source, (3) as long as the public is still freaked out / new to the idea anyone can make images and video of whatever, and (4) as long as the signing compute doesn't eat into the bottom line once everyone in the world has uniform access to the tech.

The idea here is that {law enforcement, lawyers, journalists} find a deep fake {illegal, porn, libelous, controversial} image and goes to Google to ask who made it. That only works for so long, if at all. Once everyone can do this and the lookup hit rates (or even inquiries) are It's really so you can tell journalists "we did our very best" so that they shut up and stop writing bad articles about "Google causing harm" and "Google enabling the bad guys".

We're just in the awkward phase where everyone is freaking out that you can make images of Trump wearing a bikini, Tim Cook saying he hates Apple and loves Samsung, or the South Park kids deep faking each other into silly circumstances. In ten years, this will be normal for everyone.

Writing the sentence "Dr. Phil eats a bagel" is no different than writing the prompt "Dr. Phil eats a bagel". The former has been easy to do for centuries and required the brain to do some work to visualize. Now we have tools that previsualize and get those ideas as pixels into the brain a little faster than ASCII/UTF-8 graphemes. At the end of the day, it's the same thing.

And you'll recall that various forms of written text - and indeed, speech itself - have been illegal in various times, places, and jurisdictions throughout history. You didn't insult Caesar, you didn't blaspheme the medieval church, and you don't libel in America today.

Re: Nano Banana Pro

#86

I've tried to repaint the exterior of my house. More than 20 times with very detailed prompts. I even tried to optimize it with Claude. No matter what, every time it added one, two or three extra windows to the same wall.

I also tried that in the past with poor results. I just tried it this morning with nano banana pro and it nailed it with a very short prompt: "Repaint the house white with black trim. Do not paint over brick."

Re: Nano Banana Pro

#87
Google needs to pace themselves. AI studio, Antigravity, Banana, Banana Pro, Grape Ultra, Gemini 3, etc. This information overload don't do them any good whatsoever.

Re: Nano Banana Pro

#88

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…

have some kind of standardized identifier on them Take this a step further and it'll be a personal identifying watermark (only the company can decode). Home printers already do this to some degree.

yeah, personally identifying undetectable watermarks are kindof a terrifying prospect

Re: Nano Banana Pro

#89
post #82

I've tried to repaint the exterior of my house. More than 20 times with very detailed prompts. I even tried to optimize it with Claude. No matter what, every time it added one, two or three extra windows to the same wall.

I tried this in AI studio just now with nano banana. Results: https://imgur.com/a/9II0Aip The white house was the original (random photo from Google). The prompt was "What paint color would look nice? Paint the house."

Guess they ran out of paint - notice the upper window.
Post reply on HN