This Lora can give your characters Adepta Sororitas-esque armor. Full blast for anime models, up to 0.5 for photo models. My first attempts at LoRA training, I have absolutely no idea what I'm doing. Help.
Description
Shrunk down to 9mb, lost quite a surprising amount of detail, it quite dynamic though. For more distinct and higher detail armors go with the 144mb version. Can use higher strength for photoreal models.
FAQ
Comments (16)
Do you think it's possible to do a more realistic version?
I rearranged the dataset, V2 works better on realistic checkpoints.
Let's make a male version
If you're using the 144mb versions, you can force males easily using male keywords, (male focus, 1boy, etc.) The rank8 versions seem to struggle more with that.
Isn't that called a space marine? Lol
Hello I trying to something similar to howling banshee armor, what prompt and description just to get show on any human model?
I trained the model to only describe the armor and the background, leave out description of the face, head, hair, eyes. So when you prompt you can tell it to generate a female face with the armor. Try that.
@InvictusAI Thank you
Just finding this gem today! Amazing work!
What is the difference between v2 and rank 8?
Size and a little bit of strength.
doesn't work help.
What's your workflow?
@InvictusAI Prompt: score_9, score_8_up, score_7_up, ultra realistic 8k cg, (epic:1.5, perfect artwork, masterpiece),(((detailed background))), picture-perfect face, perfect face,glowing eyes,red lips,thick eyebrows,shiny skin,intricate detail,textured hair,gorgeous goddess, deity, divine queen, delicate pattern,at ancient palace{{blunt bangs}},beautiful sultan woman,aquamarine eyes,black hair,golden jewelry,Magnificent,portrait,candles
Negative: score_6, score_5, score_4, easynegative, ng_deepnegative_v1_75t, watermark, writing, deformed, male face visible, simple background, realistic
Parameters: Steps: 30| Size: 1024x1024| Sampler: DPM2 FlowMatch| Seed: 31| CFG scale: 8| CFG rescale: 0.7| Model: autismmixSDXL_autismmixConfetti| Model hash: ac006fdd7e| Refiner negative: low quality| App: SD.Next| Version: 9ea16a3| Pipeline: StableDiffusionXLPipeline| Operations: txt2img; detailer| VAE: sdxl_vae| Detailer: GFPGANv1.4.fp16| Detailer steps: 20| Detailer strength: 0.4| Detailer prompt: detailed face, beautiful face| Detailer negative: ugly| Sampler sigma: karras
@MCStormy Ah see this is a LoRA for SD 1.5, you're using a Pony based model. I have a SDXL version available but YMMV since it's not specifically trained for Pony either.
@InvictusAI Got it, thank you.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.