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

Show HN: New AI edits images based on text instructions

github.com

Post reply on HN
Print view View original

221–230 of 241 posts

Re: Show HN: New AI edits images based on text instructions

#221

Post by YurgenJurgensen »

taberiand wrote: post #191
sandworm101 wrote: post #145

Fireworks. These AI tools seem very good at replacing textures, less so about inserting objects. They can all "add fireworks" to a picture. They know what fireworks look like and diligently insert them into "sky" part of pictures. But they don't know that fireworks are large objects far away rather than small objects up close (see the Father Ted bit on that one). So they add tiny fireworks into pictures that don't ha…

Perhaps stereoscopic video should be part of the training data?

Human stereoscopy is only good out to a few meters (and presumably people aren't going out with giant WWII stereoscopic rangefinders to generate training data). So it wouldn't help them for things like fireworks or trees.

Re: Show HN: New AI edits images based on text instructions

#222

Post by bryced »

goffi wrote: post #220
bryced wrote: post #213

Earlier quoted context omitted.

> would it be possible to use this tool to make automatic mask for editing in something like GIMP probably but GIMP plugins are not something I've looked into > REPL already done. just type `aimg` and you're good to go > GUI GUIs add a lot of complexity. Can your file manager do thumbnails and quick previews?

thanks for the quick answer and cool for REPL. Yeah sure I can just launch Gwenview on the output directory. > probably but GIMP plugins are not something I've looked into I was just thinking about a black and white or grey level output image with the desired area , no need to integrate it in GIMP of whatever. I've tried a prompt like "keep only the face", but no luck so far.

There is a smart mask feature. Add `--mask-prompt face --mask-mode keep`. I believe it outputs the masks as well

Re: Show HN: New AI edits images based on text instructions

#223

Post by seanw444 »

Zuiii wrote: post #198
sandworm101 wrote: post #115

Earlier quoted context omitted.

Nudity isnt really the core issue. It is about illegal content. Nudity is the over-protective, over-inclusive bandaid solution to prevent this thing from being used to generate the very illegal material that will trigger authorities.

Then society shouldn't tolerate knifes either. Banning an entire class of activity because someone somewhere might abuse it is a ridiculous irrational way to reason about things. The question is simple: Will the thing mostly be used for bad or good? Unless you think the vast majority of humanity are pedophiles then these features should be allowed. Think of the children was never a valid argument and isn't a valid on…

Agreed. This was an inevitable use case for AI imagery from the get-go. No way around it. Even if one dev/trainer goes out of their way to make certain that it can't be used for that purpose, another model will be trained that can. So the only three solutions are:

A. Get over it.

B. Ban the tool in its entirety.

C. Waste tons of governmental resources, spy on people more, and operate in legal grey-area to hunt down people that produce that kind of stuff with these models.

The correct answer seems obvious to me.

Re: Show HN: New AI edits images based on text instructions

#224

Post by Jerrrry »

dang wrote: post #175
bryced wrote: post #170

Earlier quoted context omitted.

If you're wanting to use Stable Diffusion 2.1 with imaginairy you just specify the model with `--model SD-2.1` when running the `aimg imagine` command.

Sorry for the offtopicness but could you please email me at hn@ycombinator.com? (I have an idea/possibility for you. Nothing that couldn't be posted here in principle but I'd rather not take the thread off topic.)

This was an uncanny comment, somehow.

Hope ya'll brainstorming session is fruitful.

Re: Show HN: New AI edits images based on text instructions

#225

Post by dang »

Jerrrry wrote: post #224
dang wrote: post #175

Earlier quoted context omitted.

Sorry for the offtopicness but could you please email me at hn@ycombinator.com? (I have an idea/possibility for you. Nothing that couldn't be posted here in principle but I'd rather not take the thread off topic.)

This was an uncanny comment, somehow. Hope ya'll brainstorming session is fruitful.

Sorry for the uncanning! My thought was simply to connect the OP with a YC partner in case they wanted to explore doing this as a startup.

I send such emails all the time but on semi-rare occasions have to resort to offtopic pleas like the GP.

I hope that helps clear things up!

Re: Show HN: New AI edits images based on text instructions

#226

Post by singhrac »

GordonS wrote: post #63

What are the most affordable GPUs that will run this? (it said it needs CUDA, min 11GB VRAM, so I guess my relatively puny 4GB 570RX isn't going to cut it!)

For what it's worth, it ran fine on my 2070 (8GB of VRAM), even with the GPU being used to render my desktop (Windows), which used another ~800MB of VRAM. I was running it under WSL, which also worked fine.

Note the level of investment that NVIDIA's software team has here: they have a separate WSL-Ubuntu installation method that takes care not to overwrite Windows drivers but installs the CUDA toolkit anyway. I expected this to be a niche, brittle process, but it was very well supported.

Re: Show HN: New AI edits images based on text instructions

#227

Post by Jerrrry »

dang wrote: post #225
Jerrrry wrote: post #224

Earlier quoted context omitted.

This was an uncanny comment, somehow. Hope ya'll brainstorming session is fruitful.

Sorry for the uncanning! My thought was simply to connect the OP with a YC partner in case they wanted to explore doing this as a startup. I send such emails all the time but on semi-rare occasions have to resort to offtopic pleas like the GP. I hope that helps clear things up!

It does, thank you!

Re: Show HN: New AI edits images based on text instructions

#228

Post by wincy »

CyanBird wrote: post #131
Daub wrote: post #92

The language of high-level art-direction can be way more complex than one might assume. I wonder how this model might cope with the following: ‘Decrease high-frequency features of background.’ ‘Increase intra-contrast of middle ground to foreground.’ ‘Increase global saturation contrast.’ ‘Increase hue spread of greens.’

They behave quite poorly, because the keywords used by the models are layman language not technical art or color correction/color grading-speak Hopefully in a couple of years when things have matured more there will be more models capable of handling said requests The most precise models are actually anime models because the users have got high standards for telling the machine what they expect of it and the database…

When I was training Dreambooth on images of myself, then trying to get tags out of the images it generated of me to write better prompts, I clicked “booru tags” on automatic1111 not knowing what it was. It thought I was some sort of Yaoi manga and generated lots of tags that made me both uncomfortable and confused.

Re: Show HN: New AI edits images based on text instructions

#229

Post by treeman79 »

YurgenJurgensen wrote: post #221
taberiand wrote: post #191

Earlier quoted context omitted.

Perhaps stereoscopic video should be part of the training data?

Human stereoscopy is only good out to a few meters (and presumably people aren't going out with giant WWII stereoscopic rangefinders to generate training data). So it wouldn't help them for things like fireworks or trees.

Found out I couldn’t see in stereo. Got prism glasses. Completely blew my mind to seen ”depth” for the 1st time. Had no idea I couldn’t. Never had any trouble.

Apparently without prism glasses my vision just switches from one eye to the next every 30 seconds. Completely seamlessly.

Re: Show HN: New AI edits images based on text instructions

#230

Post by dr_dshiv »

treeman79 wrote: post #229
YurgenJurgensen wrote: post #221

Earlier quoted context omitted.

Human stereoscopy is only good out to a few meters (and presumably people aren't going out with giant WWII stereoscopic rangefinders to generate training data). So it wouldn't help them for things like fireworks or trees.

Found out I couldn’t see in stereo. Got prism glasses. Completely blew my mind to seen ”depth” for the 1st time. Had no idea I couldn’t. Never had any trouble. Apparently without prism glasses my vision just switches from one eye to the next every 30 seconds. Completely seamlessly.

Wow! Any brand suggestions? And, are these the same prism glasses as those that let you watch tv laying in bed?
Post reply on HN