Description
Steers the model towards the style of gpt image 2. It improves details, lighting and texture a lot, though the image might end up looking slightly sloppy.
This is meant as an enhancer so play with lora strength to allow more style diversity :)
Note on versions
v1.1 is closer to the images in the dataset, it adds a lot more texture and details, shifts the colors less, and affects the pose / composition a bit less too compared to v1. It has the downside of sticking more closely to the body and face of gpt-image-2 (which gets boring fairly quickly), and having some of the less pleasing repeating texture pattern of gpt2 image.
You might just prefer the effect that v1 give if you want less texture or punchier colors.
Misc Notes
This is trained without a reg dataset, so the trigger word is not needed, but helps ever so slightly.
I found some better optimizer settings, and fixed some bug in the custom code I added to the trainer I used between v1 and v1.1. That was most likely holding back all of my anima LoRA. Which explain the large change between the two version when trained with the exact same dataset.
I might train another version since the v1.1 LoRA was still learning when the save published was auto-generated, but it started to get bad composition in the following epochs. I might need to lower the shift during training to it focuses more on the details. Though it will never be "perfect" because Qwen VAE is one of the worse VAE alongside the old SD model ones.
Description
Training Settings: batch size 4, 1024px, ~114 images per epoch, optimizer is SinkSGD_adv
I did a few multires epochs at up to 1344px after the model stabilized to bring out the details.
FAQ
Comments (8)
This is in my top 3 favorite LoRAs I’ve ever used. It is so good! :)
Best Lora created
Details should be added with restraint. Too many details can make a piece look cluttered and chaotic. For the image creator, not the LoRA creator.
This is the same dataset as the previous version funnily enough.
But I can see how you could prefer v1.
I mostly use this at 0.5 weight with other lora and it is better than the previous one for that IMO (but I also prefer v1.1 overall)
To anyone else who looked at the example images and thought it looked too messy: give the lora a try anyway. I find it meshes very well with artstyle loras without looking overdetailed and you can always drop the weight some too. 1.0 is better at keeping style. I recommend using it without the trigger word for maximum style adherence.
Good lora to add details.
I love this, thank you!
v1.1 has better graphics but follows instructions much worse than v1