I will pay for your Stable Diffusion knowledge

Hey y’all!

We’re building a product based on Stable Diffusion, and we would like to train and deploy a custom Stable Diffusion model. We’ve got some detailed questions that we can’t find answers to, so we would love to just pay for a call with someone that knows the topic. Please shoot me a DM with your hourly rate if you’re interested.

Here’s a few of the sample questions we’ve got to give some context on the expertise required:

  • How well labelled do training images need to be? Do we label backgrounds, every single thing possible on the image? Is the quality of the label proportional to the quality of the model and output images?
  • How to get the best model out, what settings to use and what GPU will we need to access? Is Google Colab “high GPU” enough?
  • Is it a good idea to feed the custom model with images produced from a base model (such as SD v1.5) and other generative AIs (such as DALLE)? This way we could easily increase our training data set.
  • Should we add examples of “ugly” images, so that later we can put negative prompting on these?

Don’t be shy and reach out!
Have a lovely day