Loading
Du hast einen ComfyUI-Workflow? Erstelle direkt im Browser eine verständliche Workflow-Dokumentation.Workflow verständlich machen.Jetzt ausprobieren ↗
Here is my third character LoRA: ❤️ KAYLA ❤️
Modellquelle
Quellenbeschreibung
Here is my third character LoRA: ❤️ KAYLA ❤️
She is an 18 year old California girl with sun-kissed blonde hair and big beautiful blue eyes.
The Ultimate California Dream!
If you use my LoRAs then please consider leaving me a tip via Ko-Fi . It helps pay for the GPU electric bills :D
Krea 2 version trained locally using AI-Toolkit on my RTX5090 with 70 specially rendered images of 3D renders of this original character on a clean background in a wide variety of poses.
Pony version trained locally using Kohya_SS on my RTX5090 with the same dataset.
Trigger word: gen3dx_kayla
Quellen
1 QuelleModellartefakte
1 Artefaktkayla_krea_v1.0.safetensors
safetensors · BF16 · 218 MB · SHA-256 bf5985077d1c…007a · Civitai
HerunterladenQuellenauszüge
1 AuszugSource context: 689 downloads · Type LORA · Base model Krea 2
For my training process, please check out my 90 minute tutorial on YouTube .
Versions:
v1.0 Krea - Kayla appears on this model in a slightly "3D/Pixar" style so you may need to adjust your Realism LoRAs if you want to get true photoreal. A couple of my samples have this look so you can see the workflow used on those.
v1.0 Pony - Superb photorealistic fidelity and very tight character consistency based on the original Daz Studio character. Trained on CyberRealisticPony checkpoint.
Some tips on usage:
Use 0.5 strength
The best likeness of her comes out on CyberRealisticPony v12.5
DPM++2M Karras is the favourite Sampler/Scheduler combo, but DPM++3M SDE, Euler, Euler A and Heun (with Realism) are also productive of good results when used with Karras or Normal scheduler.
⚠️NSFW: I have explored softcore NSFW content and it is nice and consistent, as her training set contains nudity. My warranty does not include more hardcore NSFW content. Please use this LoRA ethically!
Let me know what you think, and please share your results!
