@philosophy, I tried kindroid.ai briefly, and it is fairly enjoyable to use with minimal effort to build a character. I've not seen what it can do with images yet though I've not played with it extensively. The only noticeable tradeoff I've witnessed is it's lack of ability to finely tune responses like Kajiwoto, but that is expected given the tools I as a user have to use. I can see the appeal for someone just wanting to build a waifu with little investment. It appears a good option for most.
@DoubleD Try hitting the regenerate response button until you get a response you like. The response you pick trains the AI.
@DoubleD As far as mage.space, here are the settings I've landed on as a starting point:
Base: Realistic Vision V5
LoRAs: Varies. I often use Microwaist (deceptive name, it's just a waist size scaler) and Breast Size Slider.
Control Net: Pose ControlNet. You upload a picture, the AI figures out the pose of the pic and then applies it to what it generates.
Aspect: Phone
Steps: 50-70 (High)
Guidance: leave on default. I've played with this...it's best to leave it be.
@DoubleD mage.space also does inpainting. Inpainting is where you upload a pic, decide which parts of the pic you want the AI to redo while leaving everything else completely alone, enter your prompt, and the AI redoes those parts. For example, if you like Emma Watson but hate her flat chest, you can inpaint her chest, tell the AI to give her big ole DDs, and it (hopefully) will do it. Inpainting can take a lot of tries and practice to get right though.
