Image Generation Guide for the New Lora Based on Young Man

A newly improved model developed from  Young Man, with enhanced structural details and diverse Asian model outputs. Follow these recommended settings for the best results:

Recommended Settings

• Checkpoint: Use the new updated model based on  Young Man
• Sampler Method: DPM++ 2M Karras
• CFG Scale (Classifier-Free Guidance): 7
• VAE: vae-ft-mse-840000-ema-pruned.safetensors
• High-Resolution Upscaling: 4x-UltraSharp

Prompt Suggestions

• Clearly describe desired characteristics such as age, style, environment, lighting, and outfit details.
• For diverse Asian appearances, include terms like: "Asian woman, natural skin tone, realistic face, photorealistic, detailed textures"
• Combine with scene descriptions for better composition, for example: "Soft studio lighting, cinematic shallow depth of field, sharp facial details, vibrant background"

Additional Tips

✅ Use balanced prompts—avoid overcrowding with keywords.

✅ For extremely detailed textures, combine 4x-UltraSharp with high-resolution outputs.

✅ Adjust CFG scale between 6.5 - 7.5 depending on how strong or subtle you want prompt influence to be.

✅ Always test with a few variations to fine-tune output quality.