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

Stable Diffusion is a big deal

simonwillison.net

Post reply on HN
Print view View original

151–160 of 488 posts

Re: Stable Diffusion is a big deal

#151

Post by hackerlight »

phillipcarter wrote: post #87

I'm not really sure I know how to use these tools. I tried the following prompt: "elon musk giving donald trump a massage using pizza sauce instead of oil, in a majestic room filled with flowers and golden toilets" ...and the result was a crappy AI-generated picture of not-quite donald trump holding a terrible rendering of a pizza, and some hands sticking out of random places. There were some red flowers in the pictu…

It's only good at prompts that ask it to do a non-photorealistic artistic style. All the photorealistic attempts are bad.

Re: Stable Diffusion is a big deal

#153

Post by M4v3R »

seibelj wrote: post #50

I am a contrarian by nature. Wearing my investor hat, I continue to be unimpressed with AI tools, including the latest image generation enhancements. I do find stable diffusion interesting, but I don't see the disruption. Similarly, I don't see the disruption from Github Copilot considering I run a company full of highly paid and extremely skilled developers and not one of them uses copilot. How often have you wanted…

> I think if I was a creative writer or professional artist, the current generation of AI tools could be useful for inspiration. But the bar for quality is just so, so high in the creative industries, I am not concerned that AI will replace them.

I don't think AI will replace professional artists, but instead supplement them. Creation of Photoshop, Wacom tablets, 3d modelling and animation tools didn't put professional artists out of their jobs, it gave them new tools to create even better stuff in less time, and I think this is what's happening here too.

Re: Stable Diffusion is a big deal

#154

Post by 01100011 »

Does anyone know if it's possible, with a high end gpu or relatively cheap cloud provider, to use transfer learning to extend the model with in-house images? I think it would be useful to customize the output for certain genres or applications.

Re: Stable Diffusion is a big deal

#155

Post by paraknight »

michaelchisari wrote: post #6

After using SD heavily for a week, I half agree with this. It is incredibly disruptive, and it's wild how much it accelerates the creative process. I'll give you that. But two things I've noticed: First, artists will still have a massive advantage over non-artists with this tool. A photographer who intimately knows the different lenses and cameras and industry terms will get to a representation of their idea much fas…

I've seen an inpainting technique where you put an existing character then describe what their twin is doing, then crop out the existing, which seems to persist the character at the cost of fewer pixels to work with

Re: Stable Diffusion is a big deal

#156

Post by bulla »

michaelchisari wrote: post #6

After using SD heavily for a week, I half agree with this. It is incredibly disruptive, and it's wild how much it accelerates the creative process. I'll give you that. But two things I've noticed: First, artists will still have a massive advantage over non-artists with this tool. A photographer who intimately knows the different lenses and cameras and industry terms will get to a representation of their idea much fas…

https://textual-inversion.github.io/ Is this what you mean by the character persisting across scenarios, poses etc?

Re: Stable Diffusion is a big deal

#157

Post by machiaweliczny »

andy_ppp wrote: post #144

Is this going to start changing the world as much as search did? My guess within 10 years AI will have changed everything, it won’t look like AGI but it’ll be good enough to be better than humans at most creative endeavours including things like generating whole films and possibly games with a few carefully arranged text prompts and start images. Quite terrifying. I wonder if it will be possible to train a neural net…

IMO math > programming > rest is the order. On proof search there’s a lot of prior-work. Programs are super easy and cheap to verify or simulate/generate. BTW I bet that even if you aren’t using copilot VS Code collects your input as training data. With programs it’s not necessary to understand NL, might as well work as show me example of input->output

Re: Stable Diffusion is a big deal

#159

Post by fastball »

usgroup wrote: post #97

I've been trying to get it to draw a picture of a man trapped inside of a light bulb. Can anyone think of a prompt that works? It draws all sorts of freaky things featuring men and light bulbs but none with the former inside the latter.

https://supernotes-resources.s3.amazonaws.com/direct-uploads...

Re: Stable Diffusion is a big deal

#160

Post by amelius »

Fordec wrote: post #11
michaelchisari wrote: post #6

After using SD heavily for a week, I half agree with this. It is incredibly disruptive, and it's wild how much it accelerates the creative process. I'll give you that. But two things I've noticed: First, artists will still have a massive advantage over non-artists with this tool. A photographer who intimately knows the different lenses and cameras and industry terms will get to a representation of their idea much fas…

For every tool, a better output will always be created by people who specialize in a craft. Just as photoshop revolutionized photography, to this day you can tell the difference easily between and bad 'shops. In video games upon release everyone is bad and its an even playing field. But as people practice and gain experience they improve their usage, refine their approaches. You eventually see metas develop and best…

Yes, but art school just turned into a one-semester course.
Post reply on HN