How do you write pose directions in AI prompts to get editorial-quality fashion model results?

Editorial poses come from geometric, micro-level direction — not mood words. Specify weight distribution, hand placement, jaw and chin angle in degrees, gaze direction, and shoulder/hip line per shot. Anchor each pose to a shot type and camera angle, lock one probe image, then reuse the same constraint vocabulary as a pose library across the campaign.

Replace adjectives like "confident" or "editorial" with measurable body geometry. A working pose direction names six things: shot type, body position, weight distribution, hand and arm placement, head and gaze angle, and a single micro-expression cue. Example: "three-quarter shot, contrapposto standing, weight shifted to left hip, right hand resting on collarbone, left arm loose at side, chin tilted 15° toward camera, gaze off-frame left, lips parted slightly." That sentence renders the same way across models because every term is spatial, not emotional.

Write each pose against a category so the prompt anatomy stays consistent across the shoot. Use these five as your working set:

Standing contrapposto — "full-length, weight on right leg, left knee soft and angled in, hips rotated 10° left, shoulders square to camera, right hand at hip bone, left arm hanging straight, chin level, gaze direct."

Seated candid — "medium shot, seated on edge of bench, spine long, weight forward on right hip, right forearm resting on right thigh, left hand loose between knees, head turned 20° camera-right, gaze down and away, jaw relaxed."

Power pose — "low-angle three-quarter, feet shoulder-width, weight even, hands in pockets thumbs out, shoulders pulled back, chin lifted 10°, gaze straight into lens, mouth closed and neutral."

Walking movement — "wide shot, mid-stride, right foot planted, left foot lifted heel-first, arms in natural counter-swing, torso rotated 5° opposite hips, head facing direction of travel, eyes forward, one micro-gesture: hair lifting in wind."

Over-shoulder back shot — "medium back shot, model facing away, weight on left leg, right hand brushing back of neck, head turned 30° camera-left so jawline catches key light, gaze toward horizon."

Pair every pose with one lighting and one depth cue so the body reads sculpted — a single hard warm key light side-raked, with foreground prop / subject / midground / painted backdrop layered through the frame. Stillness instructions must include one micro-gesture (a slow exhale, a finger shift, hair lifting) — without it the model freezes stiff and the image looks dead.

Before generating a full campaign, run a probe shot: pick one full-frame image — character, garment, set, skin, pose all in it — and only scale up if it holds. In one documented editorial campaign that produced 40 stills and 30 motion clips across two models and five locations for about $150 in 630 credits, that probe gate was the difference between a clean run and burning a batch on a broken pose system.

Build your pose direction into a reusable library: 10–15 constraint-based templates covering standing, seated, power, walking, back, reclined, duo, and fragment/crop. Save them in the invideo agent's context tab so every new shoot inherits the same geometry vocabulary — that's how visual consistency holds across a campaign without rewriting pose language shot by shot. invideo is an agentic video creation tool where a creative producer agent holds project context and routes each shot to the right model — GPT-Image-2 and Nano Banana for stills, Seedance 2.0 or Kling for motion — so pose constraints written once flow through every generation.

When you need a specific reference attribute, decompose rather than dump: "take the pose from image A, the lighting from image B, the framing from image C." Hridaye, invideo's creative director, describes it as: "Just the pose. Just the framing. Just the lighting. Or a combination from all your references. Entirely conversationally: 'I want the depth structure from this one.'" That keeps the agent from blending references into mush and preserves the exact geometric nuance you wrote.

Finally, audit for editorial coverage gaps the standard pose list misses. After your base shots are locked, ask the invideo agent which shot types are missing for a "proper editorial" — back shots, product still lifes, cropped body fragments, seated reclined poses, voyeur framing, empty atmospherics, duo compositions. These pause beats are what separate a lookbook from an editorial.

Watch some of these to see what works for you:

Just the pose. Just the framing. Just the lighting. Or a combination from all your references. Entirely conversationally: 'I want the depth structure from this one.'

— Hridaye, invideo's creative director