A few months ago, I started making video clips with stable diffusion and noticed that the tools to do this were too complicated for everyday people. That's why I built neural frames. Enjoy.
Hi all, thanks for all that attention to the site. I am super happy about that. A word of caution: I can only rent 4 GPU instances currently on AWS due to service quota limits. "Unfortunately" the traffic due to hackernews is too high for that. Sorry for any inconveniences. If you are having troubles at the moment, come back later or so. I asked for service quote increase but those take usually 24h or so.
Also, very little people actually pay so I can only afford so much.
I don’t usually comment, but aws is actually rather expensive and i’ve hit this annoying quota problem as well. Been back and forth and still hasn’t been raised. I can also recommend coreweave banana.dev and pipeline.ai as great alternatives. This service looks awesome good luck with the launch!
The more AI-generated art I see the more convinced I am that it will generate entire new modes of art creation, rather than make creative work redundant.
I'm now waiting for a creation that could not have been done without AI, the labour that would be involved to create these works manually not being considered.
I've been experimenting with creating image datasets via Stable Diffusion, then training StyleGAN2, and compositing in After Effects. It's a new mode of art creation for me that I cannot recreate otherwise.
https://www.jasonfletcher.info/vjloops/
One video brought me memory of Autechre - Gantz Graf music video (warning, hard IDM style ;) [0]. It was made manually in 2002, per Wikipedia: "Rutterford also stated that there was no generative element to the imagery; every three-dimensional object in the agglomeration was painstakingly and manually synchronised with a specific element or frequency range within the track" [1].
I wonder if this could be merged with the work done to build music from inverted FFT’s to do Aphex Twin type visualization, or alternatively visualizations through oscilloscope music like Jarobeam Fenderson.
This takes me back to the demo scene days. Seeing something that’s both a technical and artistic achievement that could only have been borne out of the cutting edge. Thanks, awesome stuff.
The GPUs are rented from vast.ai
The individual machines aren't as reliable or well-integrated with other cloud services as typical cloud machines, but multiple can be put behind a queue to create a highly reliable service.
Hm, what device/os are you browsing on? The site should be mobile-friendly except for the account page. Also please feel free to reach out by email or discord.
I would like to support dreambooth, if there is a way to store the fine-tuned models more efficiently. The challenge is that each trained model is quite large and a bunch of models can't fit into one gpus memory at once.
I suggest that people simply run Stable Diffusion Deforum themselves (there's an extension for automatic1111's web UI). You can run it in a google colab notebook and the cost will likely be cheaper or the same, though I haven't bothered to compare.
1) the experience mostly works on mobile (where I first tried it). With only minimal changes in the fixed sizing I think you could make this mobile friendly.
2) for posts shared via your Twitter, would be interesting to see details about the prompt(s) used vs “new post”
3) I’d like to have a bit more customization in the options
Overall really nice and good luck with monetizing it. I’d love to see a blog post write up on the technical implementation. That’s something I’d more be willing to pay to see personally.
Those are very good suggestions. I built it with mobile in mind but somehow I failed to get it right, yet. Will continue to work on it. Would also be nice to make a native app at some point.
This is very cool! What sort of values are you using for the denoising strength/guidance scale? Each frame is a nice level of similar/different to the last to create flowing video.
I’d much prefer to run this locally on my own. Any chance you’re willing to share code (or suggestions / links since it looks like you might be trying to monetize this per your Twitter posts)?
Edit: Nevermind - I don't have autoplay active and there was no way to see that the "image" on the landing page is actually a video (no UI like a play button).
Who cares about frivolous lawsuits. Stable Diffision and the like tools will prevail. The main hope is that it will be their OSS versions and not corporate (Dall-E).
The works and ways of lives that are threatened by SD and the like are not worth preserving
I wouldn't call them frivolous. Quite a few people believe training generative AI models on their copyrighted work and providing others with access to that model violates their copyright. IANAL but it seems a weak argument to me. And it seems weak to some IP lawyers I know as well.
It also seems like technology that can't be realistically bottled back up. However, I wouldn't call lawsuits frivolous and it might actually be useful to get some legal clarity
I have yet to hear a copyright lawyer who actually understands how these tools work say anything other than that it's going to be fair use; the ones that I have seen comment on it as such seem to have extremely misunderstood how the tech works and how the images are used, they've just regurgitated the arguments that the artists opposed to it make (about how it copy and pasted images, etc).
Maybe I'm wrong, also not a lawyer, but I've heard from enough at this point (including ones that I've paid and are actually our lawyers as we evaluate internal use of the tools) that I definitely wouldn't bet on these suits succeeding.
And Google was sued many times back in the day for the same thing, because of their indexing of the web. They won out under fair use for the exact same reasons that all of these lawsuits will fail, the way these images are used falls very clearly within the dead center of fair use, and is probably even less questionable than Google's, since Google really does store and display direct copies of pieces of the content that it organizes.
I would be shocked if there were any significant tech companies that are not currently being sued. Frivolous (or maybe even not so frivolous) lawsuits are a fact of nature, best not to waste time worrying about them until you are the target.
the filing of the lawsuit shows a clear lack of understanding by those doing the suing, it will be thrown out almost immediately unless brought to a sympathetic judge, which is unlikely since they filed in the bay area.
Also, very little people actually pay so I can only afford so much.