Type | |
Stats | 421 0 3 |
Reviews | (56) |
Published | Jan 17, 2025 |
Base Model | |
Training | Steps: 8,400 Epochs: 70 |
Trigger Words | stell4r3v3 wearing bodysuit |
Hash | AutoV2 F3BA8640FB |
Hunyua
No complaints about this version, it turned out pretty great. Trained on 1024 resolution images and double blocks. The trigger word is "stell4r3v3" but you should write "stell4r3v3 wearing bodysuit" to get more accurate results.
Trained on 1024 resolution and double blocks
Flux
I am not very happy with the lora but will post it as a way to force myself to revisit it. Tried experimenting with both how bucketing works with flux, the ability to train multiple clothes and mixing 2d and 3d images. The lora works great for the "skin suit" but the "planet diving suit" is hit or miss.
I hope to revisit the lora soon but training for flux is very costly on buzz.
If you want to help me I need better high res images of either the Skin Suit or the Planet Diving Suit.
If you enjoy my work please leave a like or a review. Pics and vids of your work are also appreciated :)