Hands‑On: Training a The Witcher 3 Yennefer LoRA with Krea 2
y3kdhmbdb2ch2fc6vpm2 · reddit · 2026-07-20
The author shares practical experience training a LoRA for the character Yennefer from The Witcher 3 using OneTrainer on the Krea 2 model. The process demonstrates high controllability and detail preservation.
Training parameters and setup:
- Dataset: 76 manually cropped high‑resolution screenshots from version 4.04 (ultra quality + ray tracing).
- Hardware & software: RTX 5070 Ti 16GB + 32GB RAM, using OneTrainer.
- Core settings: Trained on Krea 2 RAW INT8 W8A8, rank 32, resolution 768, 30 epochs (2280 steps), adamw8bit optimizer, learning rate 0.0001.
- Inference: Generated with Krea2 Turbo INT8, LoRA strength 0.8‑1.0, Euler/simple sampler, 10 steps, CFG 1.0.
This LoRA supports multiple variants via different trigger words: default clothing, underwear, and nude.
More from Multimodal
- HeyGen adds a media-sourcing skill for coding agents with 75k images and 10k tracks — HeyGen · 2026-07-22
- Reddit user seeks ComfyUI NSFW text-to-image and image-to-video workflows under 20 GB VRAM — hobbyist2020 · 2026-07-22
- Krea 2 users recommend a two-pass Clownshark sampler setup for sharper image details — listopalafoto · 2026-07-22
- Gemini Omni Flash turns a boat cabin into a cave in Flow by Google — chrisfirst · 2026-07-22
- A simple workflow to turn a photo into an image prompt using Gemini, Grok, or GPT Image — harshitagu72595 · 2026-07-22
- A Reddit user proposes a consistency LoRA to keep anime and game scenes visually stable — ThirdWorldBoy21 · 2026-07-22