How do you create cinematic depth in AI fashion images without real camera depth-of-field?

Build the depth into the set, not the lens: layer every frame in four physical planes — sharp textured foreground props at 0–3 ft, the subject at 3–8 ft, a real 3D prop midground at 8–15 ft, and a deliberately soft hand-painted backdrop beyond — then bake those planes into the agent's global rules so every shot inherits them.

Start by designing each frame as four physical depth planes instead of prompting for "shallow depth of field." Foreground (0–3 ft): a sharp, textured object partially entering the frame. Subject plane (3–8 ft): your model and garment, fully sharp. Midground (8–15 ft): a real 3D prop or transition object that reads as occupied space. Backdrop: a deliberately soft, hand-painted background. The softness a lens would create is designed into the set itself — the backdrop is soft because it's painted, not because a lens blurred it.

Next, write those four planes into the agent's global rules once, so every generated image inherits the spatial structure without per-shot prompting. In the invideo agent, project context holds global rules — skin realism, posing, aspect ratio, and depth structure — and applies them across the run. One documented editorial campaign locked exactly this rule set and then produced 40 editorial stills and 30 motion clips across 2 models and 5 locations in 3–4 hours, for about 630 credits (~$150) — a workflow the team spent 2 weeks and $5,000+ testing before it held reliably.

Then add lighting separation, which reads as depth even in a still. Specify a single hard warm key light, side-raked, with the model's skin rim-lit and the environment falling into shadow — that direction produced editorial-grade subject separation in the documented campaign.

Direct the invideo agent in craft language rather than similarity prompts: "depth structure" and "raked light" produce more accurate results than "make it like this." If a reference image already has the spatial layering you want, pull just that attribute conversationally — "I want the depth structure from this one" — while taking lighting or framing from other references.

Treat the visible construction as the aesthetic, not a flaw to hide. Painted canvas backdrops with visible brushstrokes were used intentionally in the documented campaign — the constructed, theatrical look is a deliberate creative direction, so the soft plane never has to imitate photographic bokeh.

Before batching a full campaign, generate one probe shot containing character, garment, set, skin, and pose in a single full frame. If the four planes hold in that one image, the system is ready to scale; skipping it risks propagating a flat-looking frame across the entire run.

Watch some of these to see what works for you:

The set does the depth of field work, not the lens. Every shot from here on adhered to this visual direction.

— invideo's creative team