Hacker Read top | best | new | newcomments | leaders | about | bookmarklet login

Or just regular text. Or algorithms with PNG or binary blob seeds.


sort by: page size:

If only there were some automated way to convert those images to text.

couldn't you just encode the images into text?

Image extraction would be cool, but to me getting a readable block of text is more important.

Huh. Surely there's a better way to do that than post text in an image? Even just using pixels to encode the text.

I would like to also couple an algorithm to generate those images. Maybe just randomly select nouns for the title, but pass on augmented/cleaned versions of the title to some generator + CLIP to autogenerate fitting images.

Many languages support creating image maps and then writing strings to text to it. Not that hard a process.

The article shows a model that does this.

It's only a few frames, but they are entirely generated from text - no seed image or interpolation required.


Yes! I love the idea of being able to find images that actually include the text you want. Even better if the image itself is relavent.

Actually I meant have it generate a meme image from the text you provided. Like all the other meme generators out there, but much more simple.

You can poison all your images with Glaze and Nightshade. Then you don't have to stop them from using them - they have to stop themselves from using them or their image generator will be useless. I don't know if there's a comparable system for text. If there was, it would probably be noticeable to humans.

This all seems premature. So far we have something that can generate text and pictures based on other text and pictures ?

Why would you say that? Every word can be easily encoded in a tiny picture.

With modern ML, that is not true. For example, a browser could easily run OCR on the frame-buffer and read it out loud.

In fact, the approach would be more general, since a lot of text is already embedded in images on a lot of websites (think also about ads).


you simply have a tool which extracts the text, transforms it however you please, and then puts it back as an image.

Then you still need some way to handle text, and image to text is not reliable. Text simply has so much more distilled information. Images are nice for humans, but I can't imagine them as a storage format for programs.


Text rendering is more complex than decoding a PNG.

I'd like to find a way to start with an embedding and have the computer generate some text that corresponds, at least approximately. There are tools that do that for images, right? Like Stable Diffusion, you can put an image in, get an embedding, then do gradient descent in latent space to find a new embedding, then generate a new image from that.

I think this is true not just of images but also for text with things like word2vec etc

Many AI image generation can do text now. Especially some unreleased ones like Muse from Google.

What does behind the scenes might isn’t necessarily 1:1 with old image generation stuff either way.


You got that switched around. Text generation is harder than image generation.
next

Legal | privacy