This LoRA is trained to capture the distinct visual style of Korean Instagram influencers, think soft lighting, fashion-forward outfits, natural yet refined beauty, and the kind of composed, lifestyle-driven framing you'd scroll past and double-tap. Built on top of Z-Image-Base.Whether you're generating portraits, lifestyle shots, or fashion content, this LoRA pushes your outputs toward that curated, high-quality K-influencer aesthetic. To put it simply. "Brings the polished, editorial aesthetic of Korean Instagram influencers to your generations, clean skin, trendy fits, and that characteristic 'feed-worthy' look." ✦ Trigger WordUse the magic keyword to activate the LoRA style:<code>korfluencer</code> ✦ Recommended PairingPairs exceptionally well with:<a rel="ugc" href="https://civarchive.com/models/2362961?modelVersionId=2678724">Z-Image-Fun Distill LoRA</a> by UDCAI — <code>tested ver 2602-8step-distill</code>This combination gives you fast, high-quality results with strong stylistic coherence. This is good for eliminating the quirkiness of the Z-Image-Base model, if you don't want to think about negative prompts too much or increasing steps. I mainly tested the LoRA. combined with mine, with 20 steps and less. (res_multistep - denoise: 0.8-1) ✦ Version Notesv1<ul><li>310 images</li><li>Rank 64, Alpha 128</li><li>Initial release</li></ul>v1.1<ul><li>310 images</li><li>Rank 64, Alpha default (removed)</li><li>Cleaner defaults, same dataset</li></ul>v2<ul><li>590 images</li><li>Rank 128, Alpha 256</li><li>Larger dataset, higher rank, best quality and detail</li></ul> v3 - soon, planning to add more images, and increase learning rate by a bit. ✦ Usage Tips<ul><li>Start with a LoRA weight of 0.8–1.0 and adjust to taste</li><li>Works best with prompts that include fashion/lifestyle context (e.g., clothing descriptions, settings like cafés, streets, studios)</li><li>Combine with a quality upscaler for print-ready results</li></ul>

K-Influencer Instagram Style

<ul><li>590images</li><li>Rank 128, Alpha 256</li><li>Larger dataset, higher rank, best quality and detail</li></ul>

Files with SHA256: f515bbf8480e0d0ec2b19d8b213dbb606f8671aabd6548dadd21a772fee4e6de