Verified: 8 months ago
SafeTensor
The FLUX.1 [dev] Model is licensed by Black Forest Labs. Inc. under the FLUX.1 [dev] Non-Commercial License. Copyright Black Forest Labs. Inc.
IN NO EVENT SHALL BLACK FOREST LABS, INC. BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH USE OF THIS MODEL.
This is lora for female lamias trained on Flux1-dev2pro. You can choose tail color. Illustrations weren’t cherry picked (except for the titular preview). Neither were they edited in any way, aside from ADetailer. In the gallery you can find examples of cherry picked ones.
Trigger words:
lamia - needed.
woman or 1girl - recommended, as I used them in dataset captions.
(name_of_a_color) tail - recommended to avoid getting double tails, e.g. green tail, blue tail.
full body – recommended if perspective is borked.
pointy ears - if you want elf ears I suggest you write this, as lora mostly generates human ears unless specified.
Dataset info:
I used only females, full body front view, so it can do just that. No males, no R18. General style of the dataset was digital art and anime, all fantasy-related as I am going to use it in my TRPG campaign.
Quality of generation:
Tails and hands quality varies. It is enough for me, so I am putting it out like that, maybe will do some revisions in the future.
Generation parameters:
I recommend to use standard Flux parameters or of any other models based on Flux that you use. You need to specify the background if you don’t want it to be simple. Recommended lora strength: 0.7-1.0. You can find grids down below. If you want to use another loras with this one - I recommend you lower the strength.
Training parameters of lora v. 0.2: Standart for Fluxgym with enabled bucket. Model: flux1-dev2pro-single. Epochs 14, repeat trains per image 10, steps 13440, images 48 with booru-style captions.
