TL;DR prompt formula
Describe the person, pose/framing, lighting, environment, and mood/color — in that order — then add quality cues.
Formula:
[subject & identity] — [body/wardrobe] — [pose & camera] — [lighting] — [environment] — [mood/color] — [quality]
Example:
25–30 y/o Latina woman, natural freckles, chest-length wavy dark-brown hair, slim figure; white tank + jeans; 3/4 portrait, eye contact, 50mm, shallow DOF; soft window light; minimal studio backdrop; calm, neutral grade; realistic skin texture, high detail.
Negative prompts that actually help
Use a compact negative to prevent common issues, then add scene-specific bans.
Starter negative (copy-paste):
lowres, bad anatomy, extra limbs, deformed hands, fused fingers, duplicate face, blurry, over-smooth skin, harsh sharpening, watermark, text, logo, jpeg artifacts, oversaturated, extreme lens distortion
Add specifics when needed:
wet skin (if you don’t want it), see-through fabric, nudity, cartoonish, unreal glow, overdefined muscles, wide-angle distortion, overexposed highlights.
LoRAs: what changes when you use them
SDXL only. WAN ignores LoRAs.
Pick up to 3 LoRAs and set Strength per LoRA (range 0.1–3.0, default 0.8).
We auto-inject the LoRA’s keywords behind the scenes, so you don’t have to stuff the same terms into your prompt. Keep your text natural; just describe the shot.
Guidance: 0.5–1.2 is the sweet spot. >1.5 increases artefacts (warping, seams, crunchy textures). If stacking two or three LoRAs, keep each around 0.6–0.9.
Full Guide "How to use LoRas" for AI Generator by Prompt and References.
Writing for a specific person (or persona)
If you’re aiming for a real or virtual persona, be explicit. This improves identity retention and Upscale quality.
Include:
Face cues: eye color/shape, eyebrow thickness/arch, nose width/bridge, lip fullness/shape, jawline/cheekbones, facial hair or none, freckles/moles, skin tone.
Hair cues: length, texture (straight/wavy/curly), parting, color, highlights.
Body/wardrobe cues: body type, notable features (tattoo, scar), clothing fit, and fabric.
Age band & vibe: “early 30s,” “elegant mature,” “athletic editorial,” etc.
If you have a good photo of them, consider Generator by Reference or run Face Swap last on your best prompt results.
Prompt examples (people)
Photoreal headshot
Woman in her early 30s, olive skin, almond-shaped hazel eyes, soft arched brows, subtle freckles, shoulder-length dark wavy hair; studio headshot, 85mm lens, eye-level, shallow DOF; soft key + rim light, neutral grey backdrop; natural makeup; realistic skin texture, high detail.
Negative: (starter negative)
Full-body fashion
Tall slim figure, long straight brunette hair; black blazer dress + heels; full-body, 35mm, slight low angle; glossy runway set, practical spotlights; confident pose, cinematic contrast, editorial look.
Negative: bad anatomy, bent ankles, limb warping, motion blur, watermark, text
Lifestyle (plus size, curves)
Plus-size woman with soft tummy and curves, thick thighs, shoulder-length curly hair; sleeveless summer dress; candid walking pose, 50mm; golden hour park, backlight; warm filmic grade, gentle grain, realistic proportions.
Negative: cartoonish, plastic skin, oily shine, overly retouched
Athletic editorial
Fit woman with athletic tone, defined arms, high ponytail; sports bra + leggings; 3/4 pose, 70mm; studio with softbox key + kicker; crisp shadows, clean backdrop, high detail, realistic muscle definition.
Negative: overdefined muscles, unnatural veins, harsh sharpening
If you attach LoRAs like Athletic Tone or Plus Size Body, keep the prompt the same; we’ll inject the right LoRA tokens automatically.
Practical tips & hacks
Start descriptive, then prune. One tight sentence usually outperforms a paragraph.
Camera & light > adjectives. “50mm, soft window light, shallow DOF” is more reliable than “ultra realistic, 8k, masterpiece.”
One idea per run. Avoid conflicting traits (“slim” + “plus size”). If you must explore a range, run separate batches.
Use GPT as your prompt writer. Paste a reference photo (or describe it) and ask:
“Describe this image as a concise, photoreal prompt for SDXL/WAN, including camera, lighting, wardrobe, body shape, and mood. Exclude logos and text.”When realism slips:
Putting it together (copy-paste starter)
Prompt:
Early-30s woman, light olive skin, almond hazel eyes, medium-thick arched brows, straight chest-length dark hair; slim figure; white tank + jeans; 3/4 portrait, 50mm, eye contact, shallow DOF; soft window light; neutral studio wall; calm mood, subtle film look, realistic skin texture, high detail.
Negative:
lowres, bad anatomy, extra limbs, deformed hands, fused fingers, duplicate face, blurry, over-smooth skin, watermark, text, oversaturated, harsh sharpening, jpeg artifacts
Model & LoRAs:
WAN for maximum realism or SDXL + LoRAs (e.g., Classic Hourglass Shape 0.8).
If stacking LoRAs, keep each 0.6–0.9.
Troubleshooting quick fixes
Hands/limbs weird: simplify pose, add
bad handsto negative, reduce wide-angle terms, lower LoRA strengths.Face drifting between images: finish with Face Swap.
Soft details: send to Upscaler.
Outfit conflicts with LoRA: keep LoRA ≤0.8 and describe the fabric/fit clearly (e.g., “structured blazer,” “flowy cotton dress”).
That’s it — keep prompts concrete, keep negatives lean, and let LoRAs do the heavy lifting (we’ll attach their keywords for you).