A former Wall Street Journal photojournalist just explained why your production budget is about to become irrelevant.
Matthew Craig built a global visual storytelling network for brands like Airbnb and Google. Now he's running AI workflows in parallel with live shoots—and the "impossible" shots are getting cheaper than stock footage.
His prediction? The entire industry is asking the wrong question.
Here's what he shared in our latest Super Unfiltered conversation:
→ The parallel pipeline: While his team films real people, AI models train on the exact same brand guidelines. By post-production, they're synthesizing product shots that would've cost $50K and taken weeks. Clients can't tell the difference.
→ The microwave vs. the kitchen: Most creatives use AI like a microwave—type prompt, get random output. Matt's team uses Stable Diffusion like a full professional kitchen, controlling lighting, motion, climate, and environment with the precision of a DP. The gap between these approaches is the gap between amateurs and agencies in 12 months.
→ The homogenization trap: Every Midjourney output looks the same because it's trained on the statistical average of the internet. The agencies that survive will be the ones capturing their own training data during live shoots—building proprietary visual languages that can't be replicated.
The uncomfortable truth? Stock footage research used to take hours. Now generative synthesis is faster AND cheaper. The only question is whether you're still paying for the old way.
His boldest prediction for 2026: Audio. Voice becomes the primary human-computer interface, and musicians start building hyper-curated AI models. The music industry's copyright battle will make visual AI look like a warmup.
So here's Supergood's question: Are you training AI on your brand's visual language, or are you still licensing someone else's generic footage?