What does this model do? This LoRA generates realistic portraits and scenes featuring Latin American women with authentic and varied features: tanned/bronze skin, dark wavy/curly hair, expressive eyes (brown/green), and warm, sensual facial traits typical of Latin America (influences from Brazilian, Mexican, Colombian, etc.). It delivers high fidelity in details like skin texture, natural lighting, and facial expressions.What is it good for? Ideal for: - Photorealistic portraits - Female characters in stories, RPGs, or concept art - Fashion images, Latin beauty, or ethnic diversity representations - Generations in Z-Image Turbo (recommended: 4-8 steps, LoRA weight 0.6-0.8) or SDXL/Illustrious bases Main trigger words: latina, morena, latina woman (or use the specific name you defined during training, e.g., "sofia" if you captioned it that way). For best results, place the trigger at the start of your prompt.Positive prompt example: latina, morena, 25 years old, beautiful face, detailed eyes, wavy dark hair, smiling, natural lighting, photorealistic, high detail, masterpieceNegative prompt example (optional): deformed, blurry, low quality, asian, caucasian, cartoon, anime, overexposed, underexposedWhat is it NOT good for? - Anime or cartoon styles (use Pony base if you want that) - It may shift ethnicity if LoRA weight is low or you add strong conflicting prompts (e.g., "blonde hair" can force lighter features) - Avoid extreme prompts (e.g., heavy NSFW anatomy without fine-tuning) to preserve realism and avoid artifacts Trained on ~60 high-quality images using Z-Image Turbo (experimental feature on Civitai On-Site trainer). Test intermediate epochs for the best results — overtraining can cause artifacts!

Files with SHA256: 9233804afc8c2782589c4862fe86770d4bb31eca265a1dca1e588a2c96279333