
Porting my 1.5 Nobody embeds to SDXL/Pony LoRAs!
For all of these, the sweet spot for strength is 0.6-0.8; no special prompting rules, they just work.
Nico
Based on Nico Nobody; I actually ended up having two versions of his one, of different sizes. I've uploaded the smaller one here, but most of the images were made with the 64 DIM LoRA. Results should be similar; trying to upload them both in one version here.
Jada
Based on the dataset and images made with Jada Doll. As often the case, Pony can struggle with the skin color despite a larger, curated dataset and captions. So for proper use, the prompt should use:
j4d4, dark skin
Prof Jen
I got a request to see about bringing my Professor Jennifer Nobody to SDXL, and so, I endeavored to do just that! Soon followed by a Pony version.
Natalie Nobody
This one is far and away my most popular embed for 1.5 therefore I thought I'd give her life on this side of things too. Only made a Pony based training, though SDXL is only a single comment request away!
描述:
I ended up overbaking this one, so the published file is actually only Epoch 6 out of 12; 20 repeats, 4 batches, so around 1000 steps at that point.
A reminder to play around with the LoRA strength if the result looks a little strange!
I ended up overbaking this one, so the published file is actually only Epoch 6 out of 12; 20 repeats, 4 batches, so around 1000 steps at that point.
A reminder to play around with the LoRA strength if the result looks a little strange!
训练词语: profjen
名称: 494171_training_data.zip
大小 (KB): 32835
类型: Training Data
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success
名称: Prof_Jen_SDXL-000006.safetensors
大小 (KB): 223104
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success