My first Qwen-Image LoRA for the (well known) Over the mouth and hair gag.

IMHO Qwen is quite more flexible than Flux, prompt adherence is better and the native image is bigger.

Now, use the LoRA at strength from 0.5 to 1.0. See the samples for way of using. No trigger words, just a description.

No need for another bondage LoRA, this one knows perfectly well the ropes and qwen seems less censored than flux, eg it is pretty ok for tape bonds, even without a LoRA.

Civitai trainer does not support qwen (yet?) so I trained it on tensor with 100 selected images (out of 220)