Introducing Second Exposure: AI with the photographer still in the picture
Tell me if you've heard this one before.
A friend of mine is a fishing and marine photographer. His clients have started using ChatGPT to make images they used to hire him to shoot.
The tools get things wrong. The species. The equipment. Details that someone who knows the subject would catch immediately.
The clients use the images anyway. That is the problem behind my latest idea, Second Exposure.
The question I want to explore is straightforward: if a client is going to use AI for some of their images, can the photographer still be the person they hire?
Start with the photographer’s own work
Second Exposure is a project in development to build private AI image models for individual photographers, trained only on photographs they approve.
The idea is to give a photographer another way to serve an existing client. A client needs an image. The photographer uses a model trained on their work, reviews the result, and decides whether it is good enough to deliver.
The photographer keeps control over what goes out, what they charge, and whether the result belongs alongside their other work.
That last part matters. Producing something vaguely similar to a portfolio is a low bar. Whether the images reflect the photographer’s judgment well enough to be useful is something we have to test.
The boundaries are part of the idea
Asking a photographer to let you train a model on their work is asking for trust. The terms need to be clear before any files change hands.
These are the commitments behind Second Exposure:
- Photographers retain their rights to their original photographs.
- Their photographs are used only to train their own model.
- The model stays private and is used only with their approval.
- They can stop and request deletion of their images and model.
- Public use of their name, photographs, or generated results requires written permission.
The photographer also approves every image before it reaches a client. Generated images will be visibly identified as AI images, with content credentials included in the files.
An image made this way should be presented honestly. A generated fishing image cannot serve as proof of a catch, a tournament result, or an event that happened.
Those boundaries belong in the product from the beginning.
First, find out whether it works
I’m looking for photographers to help test the idea through a free pilot.
It starts with a short questionnaire about their work, their clients, and how AI is affecting their business. Before photographs are transferred or training begins, we sign an agreement covering ownership, privacy, and deletion.
Then we select photographs for training and hold others back for comparison. The photographer judges the generated results against their actual work, including images the model never saw.
If the results are useful, the next step is a small test with one client.
The pilot is planned to run about 60 days. It costs nothing; what I need in return is time, honest feedback, and a willingness to point out where the results fall short.
There is more about the process on the How it works page.
Where it could go
One direction I want to explore after the pilot is combining a photographer’s style with models of a client’s actual products: fishing lures, boats, equipment, menu items.
Getting a specific product right adds another requirement beyond making an attractive image. That work is planned for later. First, the photographer’s own model has to earn its place.
The name Second Exposure comes from photography: another exposure on the same frame. Here, the idea is new images shaped by the same eye.
I want to find out whether that can become a useful service photographers can offer their clients, on terms they choose.
If you’re a photographer seeing this shift in your own business, join the Second Exposure pilot. I’d like to hear what your clients are asking for as well as what would make this worth your time.