Qwen KILLS flux, and pretty much any other open source and most of the premium models. I think what's making people reluctant to use it is lack of LoRa support as well as a VERY FREAKING BIG diffusion model. The thing's huge. even the FP8 is huge. But its well worth the extra size, it can literally do anything. Also, there are gguf's and what-not available for it, I suggest people try it. GOTTA HAVE A GOOD PROMPT GAME, and you should get familiar with the basic prompt flow of Qwen. Its not absolutely necessary, but if you know where things should go in the prompt, it helps you get what you want with less experimentation. ANYWAY, ON TO THE LORA!
This LoRa gives people the overfilled lip filler look. Set to a strength of one and with decent prompting and trigger word, it gives very big, full lips that can only come from multiple visits to the plastic surgeon. Or, you can dial it back to simply enhance the character's lips a little. This is a very versatile model. I did not test it next to character loras but it does work with other lora and it has minimal impact on the overall look of the character even at a strength of 1. Consistent, controllable, more to come.

I'm using Qwen fp8 and Qwen bf16 along with the Qwen lightning LoRa to do most of my image generations, and the demo images were created with fp8 and lightning 8 steps. This definitely works without the lightning, and I assume it works with GGUF but I haven't tested it with that, any users who have issues with GGUF let me know and i can try to train an update.