I'm excited for this, because as an actually released model, it's possible for new tools to be built on this like this impressive collage demo. With OpenAI's DALL-E, not only could you not run it...
I'm excited for this, because as an actually released model, it's possible for new tools to be built on this like this impressive collage demo. With OpenAI's DALL-E, not only could you not run it on your own machine, you aren't even allowed to make your own tools like this for using it. There's so much more possible with AI art tools than just generating one full image from one prompt.
I would say they're intermediate in openness. From a practical perspective, a company that lets you try things out yourself without having to install anything seems more open than one that...
I would say they're intermediate in openness. From a practical perspective, a company that lets you try things out yourself without having to install anything seems more open than one that releases source code, but you need powerful hardware to run it so it's not actually useful for most people?
For image generation, Google does neither. They publish scientific papers and you can look at the canned examples. The researchers did ask on Twitter for examples that people want to try.
I briefly played around with the demo and it seems closer to midjourney than to DALL-E in quality on the all-important “can it draw an accordion” test. But maybe tweaking the parameters would help?
I briefly played around with the demo and it seems closer to midjourney than to DALL-E in quality on the all-important “can it draw an accordion” test. But maybe tweaking the parameters would help?
Yeah, it seems like Stable Diffusion is much less forgiving of "simple prompts" than DALLE (or Midjourney). This Twitter thread has some examples of good prompt construction. I've had trouble...
The terms for the beta are interesting. Users waive all rights for generated images, so anyone can use them. I ran out of free credits and bought some. Credits are sold in £10 increments, with 1p...
The terms for the beta are interesting. Users waive all rights for generated images, so anyone can use them.
All users, by use of DreamStudio Beta and the Stable Diffusion beta Discord service hereby acknowledge having read and accepted the full CC0 1.0 Universal Public Domain Dedication (available at https://creativecommons.org/publicdomain/zero/1.0/), which includes, but is not limited to, the foregoing waiver of intellectual property rights with respect to any Content. User, by use of DreamStudio Beta and the Stable Diffusion beta Discord service, acknowledges understanding that such waiver also includes waiver of any such user’s expectation and/or claim to any absolute, unconditional right to reproduce, copy, prepare derivate works, distribute, sell, perform, and/or display, as applicable, and further that any such user acknowledges no authority or right to deny permission to others to do the same with respect to the Content.
I ran out of free credits and bought some. Credits are sold in £10 increments, with 1p per "credit." Each credit generates one image at default settings (512x512).
One thing this generator can do that others can't is create reasonable (or at least recognizable) Minecraft screenshots.
I'm excited for this, because as an actually released model, it's possible for new tools to be built on this like this impressive collage demo. With OpenAI's DALL-E, not only could you not run it on your own machine, you aren't even allowed to make your own tools like this for using it. There's so much more possible with AI art tools than just generating one full image from one prompt.
Yup, OpenAI hasnt been open for a very long time.
I would say they're intermediate in openness. From a practical perspective, a company that lets you try things out yourself without having to install anything seems more open than one that releases source code, but you need powerful hardware to run it so it's not actually useful for most people?
For image generation, Google does neither. They publish scientific papers and you can look at the canned examples. The researchers did ask on Twitter for examples that people want to try.
(But at least they publish their research.)
I briefly played around with the demo and it seems closer to midjourney than to DALL-E in quality on the all-important “can it draw an accordion” test. But maybe tweaking the parameters would help?
Yeah, it seems like Stable Diffusion is much less forgiving of "simple prompts" than DALLE (or Midjourney). This Twitter thread has some examples of good prompt construction.
I've had trouble getting it to draw simple things without long and detailed prompts.
The terms for the beta are interesting. Users waive all rights for generated images, so anyone can use them.
I ran out of free credits and bought some. Credits are sold in £10 increments, with 1p per "credit." Each credit generates one image at default settings (512x512).
One thing this generator can do that others can't is create reasonable (or at least recognizable) Minecraft screenshots.