How We Teach AI to Read a Room — Scene Templates for Every Industry

A Protein Bar on a White Table Doesn't Sell Anything

When we first started generating influencer content, every image looked like it came from the same sterile photo studio. A skincare product on a marble countertop. A snack bar on a wooden table. A fitness supplement next to a shaker bottle. Technically correct, emotionally flat.

The problem wasn't the AI — it was our prompts. We were describing products without describing worlds.

Every Industry Has a Visual Language

Walk into any successful Instagram account in the food space, and you'll see warm lighting, rustic surfaces, and steam rising from dishes. Switch to fitness, and it's all dynamic poses, outdoor courts, and post-workout glow. Beauty brands live in soft focus, vanity setups, and minimalist shelving.

These aren't accidents. They're visual dialects that audiences recognize instantly. If a protein bar shows up in a cozy cafe instead of a gym, something feels off — even if the viewer can't articulate why.

Building an Industry-Aware Scene System

We built a template system that maps each industry to a complete visual vocabulary. When our system knows a business sells padel equipment, it doesn't just search for "sports" — it pulls from a curated set of settings, interactions, and props specific to that world.

A padel brand gets outdoor courts at golden hour, athletes refueling courtside, and rackets casually leaning against benches. A bakery gets warm kitchen counters, outdoor dining tables, and the act of presenting a fresh pastry to camera. A tech brand gets clean desk setups, home offices, and the ritual of unboxing.

Each template provides three dimensions: where the scene takes place, how the influencer engages with the product, and what supporting objects fill the frame. These three layers create images that feel like they were shot on location, not generated in a void.

The Cross-Contamination Problem

One of the subtler challenges was preventing scene leakage. Early on, we noticed that a food brand whose website mentioned "healthy lifestyle" would sometimes get gym backgrounds because the system matched on "health." A protein bar company would get padel court scenes because both fell under "fitness."

We solved this by adding industry-aware filtering. Food-related businesses are explicitly blocked from receiving fitness or sport templates, even if their keywords overlap. The system respects the primary industry signal and only uses secondary matches when there's no conflict.

Context Makes Products Stick

The real payoff isn't just prettier images — it's conversion. When a product appears in its natural habitat, viewers process it as a recommendation rather than an advertisement. A recovery drink held courtside after a padel match tells a story. The same drink floating in empty space is just a product shot.

Our scene templates create what we call "sticky context" — the product doesn't just appear in the image, it belongs there. The setting, the props, and the influencer's body language all work together to make the product feel like a natural part of the moment.

What's Next: Scenes That Learn

Right now, our scene templates are curated by industry. But we're exploring ways to let the system learn from engagement data — which settings drive more saves, which interactions get more comments, which props make products feel most natural. The goal is a scene system that evolves with each brand's audience, not just its industry.

Great marketing content doesn't just show a product. It places you in a moment where that product makes perfect sense.