New Workflow with LeonardoAI and an Intro to Costs of Generating AI Imagery

I’m trying a new workflow and noticed I’ve never mentioned the full history. So here is part of the EnticingAI secret sauce 😀

Besides (what I call) the Better Bodies technique. I normally use DreamshaperV7 and AbsoluteReality 1.6.

The basic workflow is like this

  1. Try a couple prompts with my regular settings
  2. Find the one I like
  3. Upscale with Smooth Upscaler (this normally gives a higher res version but no visible improvements)
  4. Download
  5. If I want a better version of the image I’d use some external upscaler to add more details.
  6. Edit in Photoshop (the amount of editing will depend on 1 and 2 heavily)
  7. Save full res and save a smaller resolution aside (I have a numbering code and the two versions allow me to play with the download settings in DeviantArt for premium downloads versions). Optional: make specific edits for Social media.

Since some days ago I gave another try to the Universal Upscaler un Leonardo.ai. I don’t know if I tired the wrong one or what but I remember it sucked and offered nothing new, also the 2x max wasn’t attractive when I can get other tools to do more. But… I’ve tried it again and notices I was getting far more better results and the 2x was probably not that bad since I could re apply the upscaler to the scaled version, thus providing me with enough resolution for a print (I normally don’t go above 5000px).

So, this is the new Workflow

  1. Try a couple prompts with my regular settings.
  2. Find the one I like.
  3. Use the universal upscaler and play with the options
  4. Upscale again of higher res
  5. Edit in Photoshop (much less than in the previous Workflow)
  6. If I want a better version of the image I’d use some external upscaler to add more details.
  7. Create extras for Social Media

So, basically it is the same number of steps but the Photoshop part is much easier, the quality is much higher.

Drawbacks and Time / Money Considerations

Business time: the Universal Upscaler takes more tokens but I don’t need to use another tool for this, so it is kind of a win.

The other issues is that my workflow depends on models that look like are going to be deprecated and the alternatives don’t provide the same features so I might also considering switching to Stable Diffusion in a self hosted environment. I think my laptop can handle it but we will see. There is also a learning curve thing I may need to consider.

LeonardoAI is a paid thing and using external Upscalers too. I see the critics on paid vs. free, the thing is nothing is free, so even having the technical ability I’m not sure if I want to sacrifice time for work, gaming and other stuff when my laptop is fully using SD + Learning Curve + Investment. I’ll stick to Leonardo as much as I can but I think it is also time to do some math and see the real time/money costs.

Hope this serves to anyone using something similar.

0 Comments

Submit a Comment

Your email address will not be published. Required fields are marked *