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

91–100 of 233 posts

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

#91

Post by quitit »

carrozo wrote: post #12
anshul wrote: post #2

Does anyone know if or when google intends to make this available for beta testing or public use?

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.”

While I understand that view, pandora’s box is already open, by their own measurements other publicly-available technology is comparable: so the only thing they’re stopping with this is direct comparison of their tech to others.

A more cynical side of me just thinks Google is rushing out the PR (including the hand-plucked sample images) because they can see that the state of the tech is progressing rapidly and perhaps by the time their tech is release-ready a competitor will already have something better (the new round of betas certainly look promising.)

It is a little bit on brand for Google to make an announcement they have the best, only for those claims to fall over later.

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

#92

Post by astrange »

MathYouF wrote: post #21
NationOfJoe wrote: post #5

i only have a passing curiosity in these projects personally. can someone in the field explain why this has exploded recently? there seems to be a lot of these tools released recently (text to image) was there a major breakthrough? a new idea that pushed everyone forward? a recent sharing of talent between groups? edit: just another thought, are they just being posted to HN now, i don't see a date on the page for whe…

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…

> I've read all the papers leading up to the latest unCLIP (Dalle2) but it's complicated. Released in 2020, with major improvements to the training process continuously being made since then.

The models behind Imagen and StableDiffusion are actually simpler than DALLE2, and both are higher quality (SD of course isn’t always since it’s much smaller). That suggests DALLE3 will also be simpler again.

There’s also been very recent work with generalized diffusion models (that use problems other than noise removal and still work) and Google researchers have been tweeting results from a merged Imagen/Parti in the last few days.

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

#93

Post by PoignardAzur »

johnthewise wrote: post #75
PoignardAzur wrote: post #64

Earlier quoted context omitted.

Interesting! I knew about transformers, CLIP and diffusion, but pixel patch encodings are new to me. Can you give me more details / point me towards an explainer? A quick duckduckgo search didn't help.

I don't quite remember whether it was first used in Vit paper[1], but it's a fairly straight forward idea. You take the patches of an image like they are words in a sentence, reduce the size of the patch(num_of_pixel x num_of_pixel) with a linear projection so that we can actually process it and get rid of sparse pixel information, add in positional encodings to put in location information of the patch and treat them…

> reduce the size of the patch(num_of_pixel x num_of_pixel) with a linear projection

What does that mean?

(Thanks for the explanation)

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

#95

Post by ShamelessC »

PeterisP wrote: post #89
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?

The writing issue demonstrably has been solved without any breakthroughs by simply making a larger model (dall-E 2 vs the publicly available dall-e), the same appears to be for other main issues as well - it's just that the publicly available versions are based on smaller/weaker models than the state of art because they're significantly cheaper to run. Also, I seem to recall that at least some models deliberately har…

Neither DALLE version 1 or 2 was completely released. Further, DALLE2 definitely still has issues with generation of text, although latent diffusion can do an okay job of it.

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

#96

Post by woojoo666 »

bmitc wrote: post #94

What is the purpose of these new AI photo generators besides crime, impersonation, or advertising? People seem really excited by them, but I don't get it.

Check out this artist on youtube that uses Midjourney to draw something in 20 minutes that would usually take him hours: https://www.youtube.com/watch?v=EsQgD9yNMxU

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

#97

Post by sheilaalien »

I'm surprised how very little this forum and the public in general knows or understands about the current capabilities and availability to these ai art generators.

You have midjourney.com beta.dreamstudio.ai craiyon.com (real quick version no fuss, low quality) creator.nightcafe.studio

and those are just some of the entry-level ones. I make ai art all day long! Check out my media feed on twitter @Sheilaaliens

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

#99

Post by astrange »

glcheetham wrote: post #66
bottlepalm wrote: post #7

Smart for Dalle, Midjourney, and Stable Diffusion to capitalize on this quickly. It looks like the technology is being commoditized at rapid speed. I wonder what’s next.

The AI is moving up the "content ladder" generating text -> now images -> next videos Smart for Google to invest in this because their business relies on third party content to exist (blogs, webpages, youtube videos) If they can vertically integrate their business to create the content AND own the discovery algorithms then they officially win the internet

Generating videos and 3D models is _much_ more difficult than images. You can’t just train off videos from the internet in the same way, because they don’t have sufficient text labels to understand them like CLIP does.

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

#100

Post by astrange »

glenneroo wrote: post #73
schappim wrote: post #30

Earlier quoted context omitted.

Is this being reported as "new news" because it is available to the public?

AFAIK nothing is released yet? They are probably still trying to, like Dall-E 1/2, remove every every image and ban every word that might generate even the slightest hint of controversial imagery before releasing it to the public. To be fair, Stable Diffusion spent months basically doing the same on Discord with thousands of beta users, and had an army of moderators flagging images which were subsequently removed in…

Even if you don’t want to block generating porn, you still want to know if you’re getting it, because nobody wants to get porn when they didn’t ask for it. (and it may be illegal depending on your country)

It’s easy to remove the filter from the SD scripts and that’s intentional.

Post reply on HN