<h1 id="yana-a-photoreal-young-woman:-short-dark-wavy-hair-with-a-soft-fringe-hazel-eyes-thick-natural-brows-full-nude-pink-lips-warm-olive-skin-hourglass-figure.-trained-on-a-20-image-studio-portrait-set-so-her-face-holds-across-poses-light-and-outfits.">Yana — a photoreal young woman: short dark wavy hair with a soft fringe, hazel eyes, thick natural brows, full nude-pink lips, warm olive skin, hourglass figure. Trained on a 20-image studio portrait set, so her face holds across poses, light and outfits.</h1>Base model: Qwen-Image-2.1SETTINGS THAT WORK (the exact values behind the gallery)- Checkpoint (UNET): qwen_image_2.1_int8_convrot — int8, weight_dtype: default- Text encoder: qwen3vl_8b_joyimage_edit_plus_int8_convrot, type qwen_image (see note below)- VAE: qwen_image_2.1_vae_bf16- Text encode node: TextEncodeQwenImage21, resolution 1024, negative prompt left empty- Sampler / scheduler: euler / sgm_uniform, denoise 1.0- Steps: 25- CFG: 1.0 — the model runs without a negative prompt here- LoRA weight: 1.0 model / 1.0 clip- Resolution: 896x1184 — ResolutionSelector 3:4 (Portrait Standard), 1 MP, multiple 8, batch 1- Speed-ups: QwenImage21Cache (prefix KV cache) + SageAttention patch (auto)All of it is straight from the workflow behind the images — nothing rounded, nothing guessed.Text encoder note. These images were made with a community fine-tune of the Qwen3-VL-8B encoder (joyimage_edit_plus). With the stock Comfy-Org encoder your result may differ slightly. Test that pairing before building a workflow around this LoRA — and tell me in the comments if the difference is big, I will retrain.WHAT SHE LOOKS LIKE IN A PROMPTportrait of a young woman, brunette, short wavy hair, bangs, hazel eyes, warm olive skin, full nude-pink lips, soft studio lighting, shallow depth of field, photorealistic, detailed skin textureHair length follows the prompt: ask for short wavy hair and you get the bob, ask for long dark hair and she grows it. Everything else in the face stays put.SAMPLE PROMPT (REPRODUCIBLE)Full-body photorealistic photograph, harsh midday sun on an open beach. The woman in her early twenties with long dark wavy hair and tanned olive skin stands square to the camera with her weight on one leg and the opposite hip pushed out, one hand lifting the wet hair off her neck and the other hanging loose at her side, chin lifted, water still beading on her shoulders and stomach. She wears a white string bikini with thin ties and a high-cut bottom, ties knotted at both hips, the fabric dark where it is wet. Behind her: blazing white sand, a hard blue sea at the horizon, one striped umbrella far out of focus. Direct overhead sun, short black shadow pooled at her feet, glare blown out on the water.seed 40251472803818 · euler + sgm_uniform · 25 steps · CFG 1.0 · denoise 1.0 · LoRA 1.0 / 1.0 · empty negative · 896x1184.EDITING WITH A REFERENCE PHOTOSame LoRA, reference mode — feed a photo of her as image 1, the prompt addresses references by index:Put the woman from image 1 into a rainy neon street at night, keep her face exactly unchanged, change her clothes to a black hoodie, cinematic lighting, reflection on wet asphaltGALLERY NOTES<h3 id="pose-light-and-outfit-variations-only-same-face-no-face-swap.-all-of-them-come-from-neyroeros.com-where-yana-has-a-full-set:-portraits-stories-and-video.">Pose, light and outfit variations only — same face, no face-swap. All of them come from <a target="_blank" rel="ugc" href="http://neyroeros.com">neyroeros.com</a>, where Yana has a full set: portraits, stories and video.</h3>LICENSEFree for personal and generated-image work. Do not resell the LoRA file or merge it into a paid checkpoint and sell that. No real identifiable people, minors, or non-consensual scenarios. Trained on Qwen-Image-2.1 (Qwen Research License).

Files with SHA256: 0687d38dbc6aaca4d1ad71ddc0f6daef7b1485b958259cb22ceeea61f9e35adf