Boreal-Qwen-Image Overview
Some attempts at a boring reality style images with Qwen Image.
These LoRAs were focused on training for very small detail, realistic lighting, and improvement in world knowledge results for the model (at least when involving people).
They do have their drawbacks comparable to SDXL with distortion settling in really quickly when you increase the lora strengths and lower the cfg.
Workflow Recommendations
I recommend combining these loras together in whatever workflow you use. Have the general discrete lora set at a strength of 1.0 and the the portrait blend high rank model to something like 0.6 - 0.8. The small low rank discrete lora may help a bit, but you need to really experiment with it.
You will need to experiment a lot to figure out the best results. I am still uncertain of the best way to work with Qwen yet and how I should approach newer versions for this style.
ComfyUI Starting Out Example Workflow: https://huggingface.co/kudzueye/boreal-qwen-image/blob/main/boreal-qwen-workflow-v1.json
Description
This version is a low rank lora trained on a smaller dataset to quickly obtain certain areas of world knowledge. It is not recommended to use this lora alone as the results may be generic and not much different than the original results from Qwen
FAQ
Comments (10)
Boss is back :)
awww yisss... just yesterday I was thinking "it would be nice to have a Qwen boreal" and here it is. Thanks!
What about wan2.1/wan2.2? That's currently best since it can do T2i.
WAN is great because it looks realistic out of the box. But if we can adapt Qwen to look real, then Qwen is superior. Prompt adherance is unheard of.
💕👍🏻
Looks good boss
How can it works with Qwen Image Edit 2509? thx!!!
this is my Amerifat model, adds 40lbs to every character lmao
does this work with Qwen Image Edit 2509?