Type | |
Stats | 73 74 |
Reviews | (10) |
Published | Jan 6, 2025 |
Base Model | |
Training | Steps: 1,600 Epochs: 6 |
Usage Tips | Clip Skip: 1 Strength: 1 |
Trigger Words | piratepunk aventgardpunk mythpunk celticpunk westernpunk ninjapunk implants enhancements mechanical_arms youngMoxie |
Hash | AutoV2 37972D747E |
Flux v2.0
This was trained 'on top of' v1.1 - IE incorporated the learning from the existing Lora (first time I have done this), and even used the same dataset. However, while I have been typically training with Flux-Dev2Pro, this was trained with Moxie Fusion Flux v1.41D. It was also trained with a lot more steps (my first 4k step Lora).
While the data sets etc. have not changed, the end result is quite drastic. It was trained with the following keywords, but I've been using it as a detailer per se with good results for scifi; piratepunk, aventgardpunk, mythpunk, celticpunk, westernpunk, ninjapunk, implants, enhancements, mechanical_arms, and youngMoxie.
v1.1 added piratepunk
v1.0
NOTE: Uploaded updated Flux 1.0 @ 2024-12-11 21:40 (added VikingPunk images to the dataset)
I'm finding the SD 3.5 Lora's I have made difficult to use and get decent images out of. However, it also gives some variety I'm not getting from Flux. Just a warning, be prepared to generate more imaged to get one you like. However w/ 3.5 Turbo, I can pop a bunch of images out a lot faster. I guess this is a warning that 3.5 is a WIP for me.
This is a combination of all the source images from all my punk styles from Moxie's Cybernetic and Punk Lora's, and Young Moxie also. It was not trained with trigger words per se, but the trigger words are part of the descriptions. Depending on the model, it may not play well with other Lora's. I'm using it at one strength and very pleased with the results.