AI girlfriend image generation: 20 lighting and pose tricks borrowed from real photographers
The vocabulary that makes real photographers good at their job is the same vocabulary that makes AI image prompts produce better results. Here are 20 terms worth stealing.
May 22, 2026 · 11 min read
Affiliate disclosure: Some of the links in this article are affiliate links. We may earn a commission if you sign up for a platform through these links, at no additional cost to you. This doesn't influence our editorial verdicts. Full disclosure →
AI image generators understand photography terminology better than they understand plain English descriptions. "Warm soft lighting from the left" produces a mediocre image. "Rembrandt lighting, golden hour, 85mm lens, shallow depth of field" produces something worth keeping. The difference is vocabulary, and the vocabulary comes from real photography.
The 20 tricks below are organized into two sets: 10 lighting terms and 10 pose/composition terms. Each one shows the photography concept, the prompt translation that AI generators respond to, and what it actually changes in the output. These work on PromptChan, Candy AI, DreamGF, OurDream, SoulGen, and self-hosted Stable Diffusion. The quality improvement from using them is immediate and consistent.
10 lighting tricks
#1: Rembrandt lighting
DRAMATIC
What it is: Named after the painter. One side of the face is lit, the other side has a triangle of light under the eye. The rest falls into shadow.
Prompt translation: "Rembrandt lighting, single light source from 45 degrees, dramatic shadow on half the face"
What changes: Adds instant mood and depth to any portrait. Transforms a flat, evenly-lit face into something with mystery and dimension. Works on both realistic and anime styles.
#2: Golden hour
WARM
What it is: The hour after sunrise or before sunset when sunlight comes in at a low angle and turns everything warm orange-gold. The single most flattering natural light for human skin.
Prompt translation: "golden hour lighting, warm sunlight, long soft shadows, amber skin tones"
What changes: Shifts the entire color palette warm. Skin glows. Shadows go soft and long. The image immediately feels cinematic rather than studio-flat. AI generators handle this term extremely well because it appears in millions of training images.
#3: Rim light / backlight
SILHOUETTE
What it is: Light coming from behind the subject, creating a glowing outline along hair and shoulders while the face stays in relative shadow.
Prompt translation: "rim lighting, backlit, glowing hair outline, soft face shadow, halo effect"
What changes: Creates a separation between the subject and the background that makes the character pop. Especially effective for NSFW content because it outlines the body while keeping some mystery in the shadows.
#4: High-key lighting
CLEAN
What it is: Bright, even lighting with minimal shadows. The look of beauty campaigns, product photography, and fashion shoots. Everything is visible, nothing is hidden.
Prompt translation: "high-key lighting, bright even illumination, white background, minimal shadows, beauty shot"
What changes: Produces clean, polished images where every detail is visible. Good for character reference shots and outfit reveals. Less atmospheric than dramatic lighting but more detail-rich.
#5: Low-key lighting
MOODY
What it is: The opposite. Mostly dark with selective highlights on specific features. Film noir, boudoir photography, and candlelit scenes all use low-key lighting.
Prompt translation: "low-key lighting, dark background, selective highlights, film noir atmosphere, deep shadows"
What changes: Hides what you want hidden and reveals what you want revealed. The most effective lighting style for boudoir and intimate scenes because the shadows do the work that explicit detail would otherwise carry.
#6: Butterfly lighting
GLAMOUR
What it is: Light source directly above and in front of the face, creating a small butterfly-shaped shadow under the nose. The classic Hollywood glamour setup from the golden age of studio portraits.
Prompt translation: "butterfly lighting, Paramount lighting, overhead beauty light, symmetrical face shadow, glamour portrait"
What changes: Produces the most conventionally flattering face lighting. Cheekbones pop, jawline defines, the face looks symmetrical and polished. AI generators respond to "butterfly lighting" directly.
#7: Neon lighting
CYBERPUNK
What it is: Colored light from neon signs casting pink, blue, and purple tones across skin and surfaces. The lighting of rain-soaked city streets and nightclub scenes.
Prompt translation: "neon lighting, pink and blue color cast, rain reflections, cyberpunk atmosphere, colorful skin highlights"
What changes: Transforms the color palette entirely. Skin picks up the neon tones, which creates a stylized look that AI generators produce exceptionally well. Overrepresented in training data, so the results are consistently strong.
#8: Overcast diffused
NATURAL
What it is: Cloud-filtered sunlight that produces soft, even illumination with no harsh shadows. The lighting that makes outdoor portraits look effortless.
Prompt translation: "overcast natural lighting, soft even illumination, no harsh shadows, cloud-filtered light, gentle and uniform"
What changes: Produces the most realistic-looking outdoor portraits. No dramatic shadows, no color distortion, just even light that shows the subject clearly. Good for character reference images where accuracy matters more than mood.
#9: Candlelight
INTIMATE
What it is: Warm, flickering, from-below light that produces moving shadows and deep orange-amber tones. The lighting of boudoir scenes, late-night conversations, and power-outage scenarios.
Prompt translation: "candlelight illumination, warm amber glow, flickering light, intimate atmosphere, soft upward shadows"
What changes: Produces the warmest, most intimate feeling of any lighting style. AI generators tend to over-orange candlelight, so adding "subtle" or "gentle" to the prompt moderates the color intensity.
#10: Catchlights
DETAIL
What it is: The small reflections of light visible in the eyes. In real photography, catchlights make eyes look alive. Without them, eyes look dead.
Prompt translation: "visible catchlights in eyes, bright eye reflections, alive expressive eyes"
What changes: Specific to close-up portraits. Adding "catchlights" to an eye-level prompt produces eyes that feel present rather than empty. Small detail, outsized impact on whether the character feels alive or mannequin.
10 pose and composition tricks
#11: Contrapposto
POSE
What it is: Weight on one leg, hips shifted, shoulders tilted opposite. The pose that every classical sculpture uses. It makes the human body look natural rather than stiff.
Prompt translation: "contrapposto pose, weight on one hip, natural asymmetric stance, relaxed confident posture"
What changes: Eliminates the T-pose stiffness that AI figures default to. The asymmetry makes the body look like it belongs to a person who's standing naturally rather than posing for a mannequin display.
#12: Rule of thirds
COMPOSITION
What it is: Placing the subject one-third from the edge rather than dead center. The most basic composition rule in photography and the one that makes the biggest difference.
Prompt translation: "rule of thirds composition, subject off-center, negative space on one side"
What changes: Produces images that feel composed rather than snapshotted. The negative space gives the image breathing room and makes the character feel like she's inside a scene rather than standing in front of a background.
#13: Shallow depth of field / Bokeh
FOCUS
What it is: Sharp subject, blurred background. Created in real photography by wide aperture lenses (f/1.4, f/1.8). Bokeh refers specifically to the quality of the blur in the out-of-focus areas.
Prompt translation: "shallow depth of field, f/1.4 aperture, creamy bokeh background, sharp focus on subject, 85mm lens"
What changes: Separates the character from the background visually, which makes the subject pop. Adding specific lens focal lengths (50mm, 85mm, 135mm) produces different amounts of background blur and perspective compression. 85mm is the portrait sweet spot.
#14: Dutch angle
TILT
What it is: Camera tilted to one side so the horizon is diagonal. Used in film for tension, unease, or visual energy.
Prompt translation: "Dutch angle, tilted camera, diagonal horizon, dynamic composition"
What changes: Adds visual energy and tension to an otherwise static pose. Use sparingly. One Dutch angle image per set is dramatic. Every image at a Dutch angle is nauseating.
#15: Low angle / worm's eye
POWER
What it is: Camera below the subject looking up. Makes the subject look taller, more powerful, more dominant.
Prompt translation: "low angle shot, looking up at subject, powerful stance, dominant perspective, worm's eye view"
What changes: Shifts the power dynamic in the image. The character towers above the viewer. Useful for dominant character types, queen/goddess archetypes, or any scene where the character should feel imposing. Pairs well with the dominatrix archetype templates.
#16: Over-the-shoulder
TEASE
What it is: Subject facing away from camera, looking back over one shoulder. The classic tease composition used in glamour and boudoir photography.
Prompt translation: "over-the-shoulder pose, looking back at camera, turned away, coy expression, back visible"
What changes: Creates a sense of catching someone in a private moment. Shows the back and shoulders while the face stays in partial view. One of the most effective NSFW compositions because it suggests without fully revealing, which makes the image more interesting than explicit alternatives.
#17: Tight crop / extreme close-up
INTIMACY
What it is: Framing that cuts in tight on a specific area: eyes, lips, hands, collarbone. Shows less but communicates more.
Prompt translation: "extreme close-up, tight crop on [eyes/lips/hands], macro detail, intimate framing"
What changes: Forces the viewer's attention to a specific detail. Tight crops on eyes produce the most emotionally charged images. Tight crops on hands or collarbones produce implied intimacy without explicit content. A useful tool for platforms with stricter content boundaries.
#18: Negative space
MOOD
What it is: Empty area in the frame that gives the subject breathing room. The nothing-there part of the image that makes the something-there part hit harder.
Prompt translation: "large negative space, minimalist composition, subject small in frame, vast empty background"
What changes: Produces editorial-feeling images where the character occupies a fraction of the frame and the emptiness creates atmosphere. Counterintuitive for NSFW content but surprisingly effective when the goal is mood over detail.
#19: Leading lines
DIRECTION
What it is: Lines in the environment (roads, railings, window frames, bed edges) that point the viewer's eye toward the subject.
Prompt translation: "leading lines composition, environment lines directing eye to subject, architectural framing"
What changes: Makes the image feel deliberately composed rather than randomly placed. The lines pull attention toward the character from the edges of the frame, which creates a sense of focus and intentionality that random compositions lack.
#20: Film grain / analog look
TEXTURE
What it is: The visible grain texture of analog film photography. Gives images an organic, non-digital feel. 35mm film has a specific grain pattern that AI generators reproduce reliably.
Prompt translation: "35mm film photography, visible film grain, analog texture, Kodak Portra 400 color palette"
What changes: Breaks the too-clean look that AI-generated images often have. The grain adds texture that makes images feel captured rather than manufactured. Specific film stock names (Portra 400, Ektar 100, Fuji 400H) produce different color palettes and grain characteristics.
Stacking terms for stronger results
The real power comes from combining terms. A single lighting term improves an image. A lighting term plus a composition term plus a lens specification produces something qualitatively different from what basic prompting achieves.
A basic prompt: "portrait of a woman in a bedroom"
The same prompt with photography vocabulary: "Rembrandt lighting portrait, 85mm f/1.4, shallow depth of field, warm candlelight fill, over-the-shoulder pose looking back at camera, film grain texture, Kodak Portra color palette"
The second prompt produces an image that looks like a professional photographed it. The vocabulary is doing the work that detailed descriptions can't. AI generators were trained on millions of captioned photographs that used these terms, which means the terms activate specific visual patterns more reliably than plain descriptions.
The 30 prompt patterns we covered earlier use similar vocabulary in the NSFW context. The 20 terms above are the vocabulary layer underneath those patterns, which you can remix in any combination the scene calls for.