My friends and I are experimenting with generating personalized outfit ideas with Dreambooth. We want to launch a site to test the idea, but building out all the infra for model training seems like an overkill. We’ve looked at https://www.astria.ai/ that @levelsio is using but their API is ridiculously expensive: $10 for a model. Plus, the quality of generations from a model trained with Astria is mediocre (e.g., many images of humans lack crucial features like hands), it takes one hour to train one model, and privacy & security info is missing; I have no idea what they do with my images.<p>Are there any other APIs we should consider?<p>Thanks!
I think you either have to deal with big prices or actually run the model yourself. There are a lot of GPU renting services that are cheap, like <a href="http://coreweave.com/" rel="nofollow">http://coreweave.com/</a> and <a href="https://vast.ai/" rel="nofollow">https://vast.ai/</a>
"Yesterday, I used a simple YouTube tutorial and a popular Google Colab notebook to fine-tune Stable Diffusion on 30 cropped 512×512 photos of me. The entire process, start to finish, took about 20 minutes and cost me about $0.40. (You can do it for free but it takes 2-3 times as long, so I paid for a faster Colab Pro GPU.)" <a href="https://waxy.org/2022/11/invasive-diffusion-how-one-unwilling-illustrator-found-herself-turned-into-an-ai-model/" rel="nofollow">https://waxy.org/2022/11/invasive-diffusion-how-one-unwillin...</a> (scroll down)