So I'm a designer trying to learn AI tools, what's the best bang for my buck to start with?

Q&A / Help / 求助答疑
1

For designers switching to AI tools, the question of what to learn first for the best ROI—my answer is start with ControlNet and img2img for control, don’t rush into flashy stuff. Reason’s simple: designers already have composition and aesthetic chops, what you’re missing is the ability to nail down the image in your head precisely.

Relying purely on prompt gacha-style re-rolling is a waste for you; what you need is to use line art, poses, reference images to constrain the model and make it obey.

Once you’ve got that down, your existing design skills can kick in immediately. As for training LoRAs or building complex workflows, learn that later when you need to mass-produce the same style.

3

Gacha-style re-rolling is really a waste for people who already have a foundation. Line art constraints are where it’s at.

4

Get good at img2img first. LoRA’s really only useful when you’re generating in bulk.

5

I got sidetracked by all those flashy workflows, it only started working once I went back and filled in the control methods.

6

“Using poses and references to constrain the model — that hits the nail on the head.”

7

One more thing, using depth maps and line art together gives you way more control.