A retrained version of my FLUX Airbag Lora with a large data set and more thorough training through tensor art.It's more consistent than the previous iteration, and will give you solid txt2img results. It's some what rigid, though, so if you want variations (particularly to the style of the car or the angle) you'll want to turn the strength down. I usually run it at 0.7.Quick Notes (I'll update as I test them):<ul><li>Does work with men or women.</li><li>Does work outside of a car.</li><li>It's biased towards having a steering wheel there. Don't know how exactly to get rid of it without img2img or inpainting.</li><li>It's pretty responsive to different textures, colors, and patterns for the airbag.</li><li>Adjectives like "giant" and "enormous" greatly effect the size of the airbag.</li><li>It does support different views and angles, but if you're doing something like a head-on shot it helps to specific where the airbags are relative to the rest of the composition (i.e. the dashboard is in front of the airbag, the airbag is in front of the subject).</li></ul>

Files with SHA256: dd773981b2c53ef5f587e80dd54fac4aff3b68f7f3530bf04d7c1cd17701d552