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

Nano Banana Pro

blog.google

Post reply on HN
Print view View original

121–130 of 718 posts

Re: Nano Banana Pro

#121

Post by jasonjmcghee »

cj wrote: post #100
jasonjmcghee wrote: post #90

Maybe I'm an obscure case, but I'm just not sure what I'd use an image generation model for. For people that use them (regularly or not), what do you use them for?

Random examples: 1) I have a tricep tendon injury and ChatGPT wants me to check my tricep reflex. I have no idea where on the elbow you're supposed to tap to trigger the reflex. 2) I'm measuring my body fat using skin fold calipers. Show me were the measurement sites are. 3) I'm going hiking. Remind me how to identify poison ivy and dangerous snakes. 4) What would I look like with a buzz cut?

First three are interesting - all question / knowledge based where the answer is a picture. Hadn't really considered this.

Re: Nano Banana Pro

#122

Post by grantpitt »

evrenesat wrote: post #73

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.

Huh, can you share a link? I tried here: https://gemini.google.com/share/e753745dfc5d

Re: Nano Banana Pro

#123

Post by swatcoder »

theoldgreybeard wrote: post #43

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…

The incentive for commercial providers to apply watermarks is so that they can safely route and classify generated content when it gets piped back in as training or reference data from the wild. That it's something that some users want is mostly secondary, although it is something they can earn some social credit for by advertising.

You're right that there will existed generated content without these watermarks, but you can bet that all the commercial providers burning $$$$ on state of the art models will gradually coalesce around some means of widespread by-default/non-optional watermarking for content they let the public generate so that they can all avoid drowning in their own filth.

Re: Nano Banana Pro

#124

Post by sib »

throwacct wrote: post #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.

Stock market seems to agree with their strategy....

Re: Nano Banana Pro

#125

Post by paulglx »

cj wrote: post #100
jasonjmcghee wrote: post #90

Maybe I'm an obscure case, but I'm just not sure what I'd use an image generation model for. For people that use them (regularly or not), what do you use them for?

Random examples: 1) I have a tricep tendon injury and ChatGPT wants me to check my tricep reflex. I have no idea where on the elbow you're supposed to tap to trigger the reflex. 2) I'm measuring my body fat using skin fold calipers. Show me were the measurement sites are. 3) I'm going hiking. Remind me how to identify poison ivy and dangerous snakes. 4) What would I look like with a buzz cut?

You should never rely on AI to do 1, 2 or 3, especially a sloppy model like this.

Re: Nano Banana Pro

#126

Post by nwsm »

crazygringo wrote: post #113
throwacct wrote: post #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.

Why? They're mostly different markets. Most people using Nano Banana Pro aren't using Antigravity. A cluster of launches reinforces the idea that Google is growing and leading in a bunch of areas. In other words, if it's having so many successes it feels like overload, that's an excellent narrative. It's not like it's going to prevent people from using the tools.

Google will never beat the "sunset after 2 years" allegations on all products that don't have "Google __" in the name

Re: Nano Banana Pro

#127

Post by evrenesat »

grantpitt wrote: post #122
evrenesat wrote: post #73

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.

Huh, can you share a link? I tried here: https://gemini.google.com/share/e753745dfc5d

https://gemini.google.com/share/79fe1a38e440

Re: Nano Banana Pro

#128

Post by ZeroCool2u »

xnx wrote: post #115
ZeroCool2u wrote: post #48

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…

You might try it again with style transfer: 1 image of style to apply to 1 target image

This is a good idea, will give it a try!

Re: Nano Banana Pro

#130

Post by vunderba »

dangoodmanUT wrote: post #45

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!)

I'd be curious about how well the inline verification works - an easy example is to have it generate a 9-pointed star, a classic example that many SOTA models have difficulties with.

In the past, I've deliberately stuck a Vision-language model in a REPL with a loop running against generative models to try to have it verify/try again because of this exact issue.

EDIT: Just tested it in Gemini - it either didn't use a VLM to actually look at the finished image or the VLM itself failed.

Output:

  I have finished cross-referencing the image against the user's specific requests. The primary focus was on confirming that the number of points on the star precisely matched the requested nine. I observed a clear visual representation of a gold-colored star with the exact point count that the user specified, confirming a complete and precise match.
Result:

  Bog standard star with *TEN POINTS*.
Post reply on HN