Input Is All That Matters

Sam Kaplan

July 12, 2026

Last week I took a photo of friends at dinner. It was bad. Tilted, focus on the wrong table, half the group in shadow. Ten years ago, that photo goes in the trash.

Instead I ran it through an image model and got the photo I wish I'd taken. Perfectly framed, perfectly lit, and still them - same people, same table, same moment. Nothing invented, everything fixed.

That bad photo wasn't a failed photograph. It was a perfect input.

For a hundred years, the craft of photography lived in the moment of capture. Framing, light, focus, timing - you had one shot to get it right, and getting it right was the art. That era is ending. Your phone's job is no longer to produce a finished image. Its job is to collect data about a moment: who was there, where, what happened. The model turns that data into the finished thing. The camera is becoming a sensor.

Video makes this even clearer.

Say you and your friends film a sketch this weekend. The bit is genuinely funny. Doesn't matter - it will look amateur, even on the newest iPhone. The lighting is whatever the room gives you. The angles are wherever someone could stand. The mic picks up the fridge.

But everything valuable is already in that footage. The banter, the timing, the performances. That's the input. The rest is production, and production is exactly what models are about to absorb.

Today, images are basically solved. Video is close: models can already generate stunning footage, and the frontier is reliably preserving your specific people, performances, and audio across a full scene. That gap is closing fast. Seedance 2.5, releasing next week, alone shows it.

Once it closes, you film your sketch and the model handles everything else. Clean audio, cinematic framing, lighting that looks like a crew spent a day on it. You could shoot in a sewer, and if the scene calls for a penthouse, you get a penthouse. What your phone captured wasn't footage. It was a performance, and the model builds the film around it.

So here's the shift: in the real world, we're moving from creating to collecting. The finished work gets made by a model. Our job is to gather the raw material only we can gather - and that material becomes more valuable, not less. Anyone will be able to generate a generic beautiful video. Only you can capture your friend's actual delivery of the joke. Real moments become the scarce ingredient.

Play it forward and essentially every image and video you see will have passed through a model. Not conjured from nothing - produced from real inputs. That's a spectacularly good deal for anyone with something to say, because the bottleneck was never ideas. Every friend group has bits funnier than half of what's on TV. What they lacked was a production team. Now everyone gets one.

Picture 2028: you and three friends spend a Saturday goofing around with an idea, phones propped on chairs, no lights, no mics. By dinner you're watching something that looks like it cost a studio a million dollars, and it's unmistakably, specifically you.

Don't worry about getting the shot right. Just get the shot. The input is all that matters.