How we use AI to generate ad creatives at scale

Product photography for D2C brands is expensive and slow. A traditional shoot with a photographer, models, and a studio takes days to plan, hours to execute, and delivers a fixed set of images. And you need fresh creatives every few weeks to fight creative fatigue.

We built an AI creative production system to solve this. It generates product photography and video content that's good enough to test, and we only invest in professional production for the winners.

How it works

The pipeline has two stages:

Stage 1: Image generation. You give it a product. It brainstorms multiple completely different "shots": editorial setups, lifestyle contexts, festive scenarios, minimalist compositions. Each shot gets a detailed multi-layer prompt describing the scene, lighting, mood, and product placement. Then it generates the images using the best available AI models.

Stage 2: Video generation. Take any generated image and turn it into a short video clip. Camera push-ins, slow reveals, fabric movement. The kind of clips that work as the opening frame of a Meta video ad.

The brainstorm-first approach

The key insight was: don't just generate "a photo of the product." Generate radically different visual worlds for the product to exist in. A silk saree in a palace courtyard at golden hour. The same saree draped over a modern chair in a minimalist apartment. The same saree on a model walking through a flower market.

Variety is what fights creative fatigue. If every ad looks like the same product-on-white-background, your audience tunes out in a week. Multiple different visual worlds give you multiple different ad angles to test.

Why this matters for performance marketing

The economics of AI creative production change the testing game fundamentally. Instead of shooting 10 images and hoping 2 work, you can generate 50 variations and test them systematically. The cost per variation is a fraction of traditional production, which means you can test more, learn faster, and find winners before your competitors do.

We use this system across our D2C brands, primarily for ethnic wear and fashion, categories where visual variety matters enormously. The production quality has reached a point where AI-generated test creatives perform comparably to studio photography in A/B tests. The studio shots still win for hero content, but for ad testing, AI is faster and more versatile.