Zien is a realistic LoRA trained entirely on AI-generated images using a realism-focused pipeline. No real people were used or referenced in training.
Use with a weight of 0.7 to 0.8 depending on prompt complexity and desired intensity.
Recommended Settings:
Prompt: natural and simple (avoid excessive keywords)
Sampler: DPM++ 2M Karras or SDE
Steps: 20–30
CFG Scale: 5–6
Resolution: 512x768 (ideal for upscaling later)
Hires Fix: 2x upscale, 10–15 steps, denoise 0.3–0.35
Upscaler: 4x_NMKD-Superscale-SP_178000_G
Description
This model is entirely trained on AI-generated images from realistic vision datasets. No real people were used or represented in the training process.