avaka-viveka-dancerCharacter LoRA for a classical Indian temple dancer. Use the trigger word in every prompt. Keep the rest of the prompt short.<h2 id="1.-quick-start">1. Quick start</h2><ul><li>Load your base checkpoint first SDXL (any version)</li><li> Add the LoRA. Start at weight 0.7–0.85. Raise only if the face or costume fades.</li><li>Put the trigger first in the positive prompt.</li><li>Use a full-body or mid-shot framing so limbs stay readable.</li><li>Keep negatives lean. Extra limbs and blur are the main failure modes.</li></ul><h2 id="2.-trigger-and-weights">2. Trigger and weights</h2>ItemValueTrigger wordavaka-viveka-dancerRecommended weight0.70 – 0.85 (try 1.0 only on a weak base)Clip skip1 for SDXL/Flux · 2 if the base is an anime/Illustrious modelResolution1024×1024 or 768×1280 portrait. Avoid tiny crops.<h2 id="3.-prompt-recipe">3. Prompt recipe</h2>Positive (keep it short):avaka-viveka-dancer, sitting in templeNegative:blurry, multi-limbed, extra arms, extra legs, extra fingers, deformed hands, cropped head, lowres, watermarkFormula: trigger + one pose + one place. Do not stack five style tags on the first try.

avaka-viveka-dancer SDXL

avaka-viveka-dancerCharacter LoRA for a classical Indian temple dancer. Use the trigger word in every prompt. Keep the rest of the prompt short.1. Quick start<ul><li>Load your base checkpoint first (SDXL any version)</li><li>Add the LoRA. Start at weight 0.7–0.85. Raise only if the face or costume fades.</li><li>Put the trigger first in the positive prompt.</li><li>Use a full-body or mid-shot framing so limbs stay readable.</li><li>Keep negatives lean. Extra limbs and blur are the main failure modes.</li></ul>2. Trigger and weightsItemValueTrigger wordavaka-viveka-dancerRecommended weight0.70 – 0.85 (try 1.0 only on a weak base)Clip skip1 for SDXL/Flux · 2 if the base is an anime/Illustrious modelResolution1024×1024 or 768×1280 portrait. Avoid tiny crops.3. Prompt recipePositive (keep it short):avaka-viveka-dancer, sitting in templeNegative:blurry, multi-limbed, extra arms, extra legs, extra fingers, deformed hands, cropped head, lowres, watermarkFormula: trigger + one pose + one place. Do not stack five style tags on the first try.

Files with SHA256: c5c964fbf515c3424e9e00327a4be973efdec29c514020aa584a8a069e246364