The Lora model is trained based on the SDXL base model
Using BLIP as a Tag translator, the first attempt (the effect is not sure), I suggest that you use natural language to describe the screen
Trigger words: mrares, bookstore
Suggested range of weight: 0.6-1.0 is acceptable, the recommended value is 0.8, the specific value should be adjusted according to the scene!
Sampler: use what you want
Use carrier: It is recommended to run the map on comfyui, and the hardware can be used on WebUi
Number of sampling steps: 25 starts, no cap
It is recommended to turn on the refiner, and the number of steps is 20 steps
————————————————————————————————————————
Here are a few example pictures (the specific prompt words will be provided to you on the example picture):
Added Lora weight: 0.8
Without Lora, the SDXL bottom mold is straight out:
————————————————————————————————————————
Added Lora weight: 1.0
Without Lora, the SDXL bottom mold is straight out:
If you want to know more, you can contact
you can send an E-mail to: [email protected]
(please include your intention, thank you)
Description
This version is the Lora model of the interior design of the modern bookstore mainly trained by SDXL as the base model, and there are also a small number of outdoor scenes (if you are interested, you can try it)
I tried to use BLIP as the translation of Tag files for the first time. I was a little lazy in the marking process, and opened the model download. If you are interested, you can try it locally.
Trigger words: mrares, bookstore