Type | |
Stats | 6,378 59,691 |
Reviews | |
Published | Sep 5, 2024 |
Base Model | |
Training | Steps: 5,000 |
Usage Tips | Strength: 0.8 |
Trigger Words | photo |
Hash | AutoV2 566C00ED0A |
Update 09/05/24: Uploaded a new version of Boreal-FD.
This version includes a fix on the latent shift dot issue, while still using the better trainer version that the older model used. This model has less creativity in areas due to its smaller dataset, but it behaves alright at much lower lora strength and does a decent job at flash photography.
I am not very satisfied with how the v2 lora out came and am going to continue training a better version for it. I had been going back in forth over the past few weeks constantly tweaking training parameters across multiple trainers and datasets. I ended up using a smaller balanced dataset with single word captions on an older commit for the AI-Toolkit while update the late shift error.
Notes on the older versions below here.
Work in Progress This is a very early experimental lora for Flux-Dev. It uses the Boring Reality image datasets to work on shifting Flux towards more realistic images.
Do not expect good image generations from them. They will be replaced by newer versions has better understanding on how to train Flux Models Emerges. The ideal workflow to use with them has not been fully understood yet either.
Additional Notes These two flux-dev loras are not expected to create very good images. Many results may be overfitted, distorted, and have this slight faded dotted look for lesser known concepts.
The 1000 step lora is more over-fitted with with distortion and lack of prompt understanding more likely to occur, but it may perform better on things like dynamic posing and skin texture.
You will want to experiment between the two loras, tweaking the lora strengths between 0.5-2.0 and guidance between 3.0-5.0 along with testing many different seeds.
As more understanding develops for Flux, better workflows for these current models will come along as well as newer Boreal-FD versions as the training improves.