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

Imagen: An AI system that creates photorealistic images from input text

imagen.research.google

Post reply on HN
Print view View original

51–60 of 233 posts

Re: Imagen: An AI system that creates photorealistic images from input text

#51

Post by ShamelessC »

boppo1 wrote: post #46

Why is this text-based image research advancing so rapidly? Is there a market application they're aiming for? Seems like multiple teams have been on multiple models and I see a new one every week.

It’s basically the first time we’ve automated perception of vision as it relates semantically with language.

In terms of theory, such systems are candidates for a generic perception engine you might use in say, a robot with cameras, speaker and a microphone.

Perception is just one aspect of intelligence, but this research ultimately makes it possible for a machine to encode data semantically.

Re: Imagen: An AI system that creates photorealistic images from input text

#52

Post by keizo »

fvold wrote: post #31
mshenfield wrote: post #28

These tools are amazing for prototyping. I had an idea for a promotional poster, and seeing my idea just by writing it felt like magic. The generated image had too many artifacts to use, but gave me a guideline to follow when creating the real thing in Pixlr. AI content generation (text, image, source code, video, music) will be a huge boon for prototyping where applied judiciously.

Yep, I feel that exact way about Nvidia Canvas [1]. It does not produce anything even close to usable as a final product, but it can produce an amazing start to a concept. [1] https://www.nvidia.com/en-us/studio/canvas/

wow, that's kinda insane and almost looks more fun for those of us short on words.

Re: Imagen: An AI system that creates photorealistic images from input text

#53

Post by tiborsaas »

boppo1 wrote: post #46

Why is this text-based image research advancing so rapidly? Is there a market application they're aiming for? Seems like multiple teams have been on multiple models and I see a new one every week.

It's like a love child of Photoshop and Polaroids with the simplicity of writing text. It would have been amazing if it didn't became popular.

Re: Imagen: An AI system that creates photorealistic images from input text

#54

Post by IvanK_net »

hackernewds wrote: post #39
mshenfield wrote: post #28

These tools are amazing for prototyping. I had an idea for a promotional poster, and seeing my idea just by writing it felt like magic. The generated image had too many artifacts to use, but gave me a guideline to follow when creating the real thing in Pixlr. AI content generation (text, image, source code, video, music) will be a huge boon for prototyping where applied judiciously.

> Creating the real thing in Pixlr is it that good now? note that Pixlr was bought by Google

Pixlr was not bought by Google. So many people spreading lies on the internet :(

Re: Imagen: An AI system that creates photorealistic images from input text

#55

Post by zo1 »

Traubenfuchs wrote: post #32
carrozo wrote: post #12

Earlier quoted context omitted.

Towards the bottom of the page they say: “The potential risks of misuse raise concerns regarding responsible open-sourcing of code and demos. At this time we have decided not to release code or a public demo. In future work we will explore a framework for responsible externalization that balances the value of external auditing with the risks of unrestricted open-access.”

Are they afraid of lawsuits or are they painfully regressive, prude moralists? Stable diffusion was un-neutered within 24 hours of its public release and the worst people do with it is Emma Watson porn.

In a nutshell they're afraid of people using prompts generating black people looking like monkeys/gorillas. And other such sensitive examples that others have posted/mentioned.

Re: Imagen: An AI system that creates photorealistic images from input text

#56

Post by andrewshadura »

curiousssnake wrote: post #42

There are some obvious mistakes in tools like this. Such as: Human faces are wrong, writing is usually scrambled, fingers look weird etc... Do you know if we need to have a major breakthrough similar to what happened 6 months ago to fix this or could these be fixed with incremental improvements in current techniques / datasets?

On the other hand, those are all things humans have troubles with when drawing, unless they have an extraordinary talent or a lot of experience.

Re: Imagen: An AI system that creates photorealistic images from input text

#57

Post by echelon »

mshenfield wrote: post #28

These tools are amazing for prototyping. I had an idea for a promotional poster, and seeing my idea just by writing it felt like magic. The generated image had too many artifacts to use, but gave me a guideline to follow when creating the real thing in Pixlr. AI content generation (text, image, source code, video, music) will be a huge boon for prototyping where applied judiciously.

Google hasn't released squat.

Google's product is vaporware and we shouldn't afford them any airtime until they release something usable. They're just trying to butt in and get press off of the backs of the teams actually working in the open, and that's super lame.

Release your model, Google, or stop bragging and talking over the others here. You're greedily sucking oxygen out of the conversation, and as a trillion dollar monopoly you don't deserve anything for free off of the backs of others. Not when you're not contributing. Stop being the rich kid talking over everyone else about how awesome your toys are.

Anyhow, the real story is Stable Diffusion. They're actively demonstrating the correct way to run this as opposed to the entirely closed OpenAI DALL-E or the (again vaporware) Google non-product.

Even MidJourney uses Stable Diffusion under the hood, using sophisticated prompt engineering to make their product distinct and powerful.

Re: Imagen: An AI system that creates photorealistic images from input text

#58

Post by MathYouF »

themoonisachees wrote: post #44
MathYouF wrote: post #21

Earlier quoted context omitted.

Like anyone deeply in a field I know maybe several thousand people who could probably give a better answer, but I figure I'll give an effort to provide one since I don't see any good ones posted yet. The moment everyone knew this was going to be big was in 2019 when StyleGAN came out. They used a lot of tricks like aligning face features (like eyes) and had all their pictures of a single domain (the most famous being…

Is 3d a different problem, or a similar one but considerably harder? I'd expect the data encoding (vertices vs pixels) to change a bit about it but I'm not familiar enough to know.

Pixel values are discrete (length x width x r256 x g256 x b256) and vertex values are continuous, so that is one major difference.

Secondly, there's vastly more labeled image data in the world than 3D data, so creating a CLMP (contrastive language and mesh pairing) model is harder.

It's very late but I may be able to give a much better answer on more of the nuances of 3D generation tomorrow.

Post reply on HN