The fastest and cheapest way for you to do this while you're "testing the waters" is to do it manually. As long as you know that the idea can be somewhat easily scaled by software in the future (it can), it's fine and encouraged to do initial tests of the product's viability with a manual implementation, in which you're doing the grunt work of finding photos yourself. Doing it that way will not only let you get started faster and be cheaper, it will also let you start developing a very fine grained feeling and understanding of why it works when it does (e.g. why person A decides to pay for their edited photo, but person B doesn't).

Once you've gotten evidence that people will pay for the service, and you have gained a better understanding of the subtleties behind how to best implement it (e.g. how to target more people like 'person A'), then you can go about spending more time / money on developing an automated bot to carry out the service at a more scalable level.

if you'd like more suggestions based on the specifics of your product idea, let me know,

best,

Lee