Summary
Apply the same techniques a photography art director uses on set, lighting, framing, focus, and finish, to get richer, more premium, more human results from Shoots. This article builds on the basics in How to prompt Shoots.

Art direction means describing a photograph the way a photographer would set it up on location, not just naming the subject. A one-line prompt like "our serum bottle on a marble shelf" gives Shoots very little to work with, so it fills in the gaps with generic choices. Adding the same details a photographer would specify, where the light comes from, how the shot is framed, what the surfaces look like, gives Shoots enough information to produce a specific, premium, human-looking result instead of a generic one.

You don't need every technique below in every prompt. Two or three, chosen to suit the shot, is usually enough.

Image prompt #1:
@Arc Accent Chair in a living room.

Image prompt #2:

Hyperrealistic editorial photo of a mid-century modern conversation pit: walnut panelling, travertine floors, oversized linen sofas in earthy neutrals, timber coffee tables, floor-to-ceiling glazing. 1970s California modernist architecture — brutalist concrete, timber, stone, indoor greenery. An @Arc Accent Chair sits naturally at the focal point among design books, ceramics and sculptural objects. Long afternoon light casts dramatic geometric shadows and warm reflections off timber. Lived-in, calm, expensive — not showroom perfect.

Camera angle and focal length refers to how the shot was taken. The table below covers the terms that make the most difference.

Example: "Shot at eye level with a shallow depth of field, so the background blurs softly behind the product."

Lighting is the single biggest lever for mood in a prompt. Rather than naming a mood directly, such as "moody" or "bright," describe the physical light source and what it is doing to the scene.

For example:

Material and texture describe how a surface physically behaves, and naming this gets more accurate results than naming the material alone. Instead of "a glass bottle," describe how it reacts to its environment: "condensation beading on a chilled glass bottle." Instead of "a wooden shelf," describe the surface: "the warm honey grain of the timber shelf."

A small amount of imperfection is what makes an image read as a real photograph rather than a computer generated one. Add one or two of the following to a prompt, not all of them:

⭐ Important: use these details sparingly

One or two finishing details per prompt is enough. Stacking more than that makes the image look like it has a filter applied, rather than looking like a photograph.

AI generated imagery can sometimes have an overly perfect, synthetic quality that makes an image read as computer generated. Counter it by asking for natural asymmetry, unretouched skin texture, and slightly imperfect framing, and by explicitly ruling out a perfectly symmetrical arrangement, glossy or plastic-looking skin, and hyper-sharp focus across the entire frame.

Example: "Natural asymmetry in the styling, unretouched skin texture, avoid a perfectly symmetrical arrangement or overly sharp focus across the whole frame."

Putting it together means combining framing, camera language, lighting, material, and finish into a single prompt, rather than using them as a checklist. The example below shows the same brief written as a basic prompt, then as an art-directed one.

Basic: "Create a clean lifestyle shot of our oat milk on a marble kitchen bench, soft morning window light, coffee accessories nearby, this image is for an advertisement."

Art directed: "A lifestyle shot of our oat milk carton on a marble kitchen bench, shot at eye level with a shallow depth of field so the coffee accessories in the background blur softly. Warm morning window light grazes across the carton from the left, catching condensation on its surface. Finished with fine film grain and a soft vignette toward the edges, natural asymmetry in the styling, this image is for an advertisement."

The second version keeps the same subject and setting, and adds camera behaviour, physical light, material detail, and a restrained finish. That is the only difference, and it is what moves the result from generic to premium.

Use this table as a shorthand when you're building a prompt. Pick one or two phrases from each row that fit the shot.

If a generated image isn't quite right, here are common visual issues and how to fix them with an edit.