SD1.5 Real Portraits LoRA V1

This is a custom LoRA trained on 11 carefully selected reference images to capture the facial features and overall appearance of
 
 Janka
 
 .
 
 The training was performed using Stable Diffusion 1.5 as the base model, with network dimensions set to 16 and network alpha to 8 for optimal balance between detail and generalization.


 
 Recommended settings:
 

<ul>
 <li>
 
 
 Base model:
 
 SD (Stable Diffusion 1.5)
 
 </li>
 <li>
 
 
 Data type:
 
 fp16
 
 </li>
 <li>
 
 
 Trigger word:
 
 <code>
 Janka
 </code>
 
 </li>
 <li>
 
 
 Suggested prompt:
 
 
 
 A high-quality portrait of Janka, ultra-detailed, sharp focus, soft lighting, 8k resolution
 
 
 </li>
 <li>
 
 
 Steps:
 
 20–30
 
 </li>
 <li>
 
 
 Sampler:
 
 Euler or DPM++ 2M Karras
 
 </li>
 <li>
 
 
 CFG Scale:
 
 7–9
 
 </li>
</ul>

 
 Notes:
 

<ul>
 <li>
 
 Using the trigger word
 <code>
 Janka
 </code>
 ensures the model activates the trained style and likeness.
 
 </li>
 <li>
 
 Works best in portrait photography prompts but can be adapted to other settings (fantasy, casual, artistic).
 
 </li>
 <li>
 
 For best results, combine with detailed background descriptions and good lighting prompts.
 
 </li>
</ul>