AI Portrait Prompts: How to Generate Realistic and Stylised Characters
- Aug 25
- 2 min read
Portrait generation is one of the most compelling areas of AI art — and one of the most technically demanding. When the subject is a human face, every viewer brings powerful pattern-recognition instincts to the image. We notice subtle wrongness immediately. We respond to expression, to light on skin, to the quality of a gaze.
Why portraits are technically challenging
AI models have improved dramatically at generating faces, but the challenge remains: the human visual system is extraordinarily sensitive to faces. Common issues: eyes that are slightly asymmetric or lack depth, hands and fingers (notoriously difficult for AI to render correctly), skin textures that look slightly plastic, and expressions that feel slightly 'off' without being obviously wrong.
The core portrait prompt vocabulary
Photography terms that improve realism: 85mm lens (the classic portrait focal length), 50mm lens (natural, close to how the eye sees), 35mm lens (slightly wider), depth of field, shallow focus, bokeh background, natural light, window light, studio lighting, shot on Canon EOS R5 or Nikon D850 (camera model references push realism).
Lighting for portraits: Rembrandt lighting (warm light on one side, triangle of light under one eye), soft natural light (gentle, flattering), dramatic side lighting (creates strong shadow contrast), golden hour (warm, glowing skin tones), backlit or rim light (creates a glowing halo effect), studio strobe or ring light (commercial/beauty photography feel).
Expression and character: contemplative (eyes looking away from camera), direct gaze (confident), and mood words like melancholy, wistful, joyful, serene that influence expression.
Negative prompts for portraits
Portraits benefit greatly from targeted negative prompts. Add: 'extra fingers, distorted hands, asymmetric eyes, plastic skin, overly smooth, doll-like, watermark, text, blurry.' For Midjourney: --no extra fingers, distorted hands, asymmetric face, plastic skin, text, watermark
Photorealistic portrait: step by step
Starting point: 'a woman.' Add age and character: 'a woman in her mid-40s with silver-streaked hair and warm eyes.' Add context: 'sitting at a wooden desk with books behind her.' Add photography language: '85mm lens, natural window light, depth of field, photorealistic.' Add lighting detail: 'soft shadows, warm skin tones, detailed skin texture.' That final prompt will produce a nuanced, characterful portrait.
Stylised and painterly portraits
Oil painting portrait: 'an elderly fisherman with a weathered face and bright eyes, oil painting, Sargent-style brushwork, Rembrandt lighting, rich dark background, painterly, museum quality.' Watercolour portrait: 'a young woman with curly hair and freckles, watercolour illustration, soft washes of colour, warm palette, airy.' Fantasy character: 'a battle-worn female knight in ornate armour, concept art, dynamic lighting, detailed, ArtStation quality, epic fantasy.'
Fantasy and stylised characters
For character portraits, the combination of cinematic style with concept art vocabulary produces particularly strong results. Example: 'a cyberpunk woman with neon tattoos and augmented eyes, rain-soaked city background, teal and orange colour grade, cinematic, close-up, ultra-detailed.' For the full styles context, see The Ultimate Guide to AI Art Styles.
Part of our AI Art Styles series:
Comments