Jason Antic

For deoldify, I think to a large extent, the research is the product. I don't see another competitive moat there, at least yet. So giving away the research seems like a dangerous step.

Yes, this is exactly how I see it. Don't get me wrong....I'd like to show everyone what was done and in fact I've had a lot of back and forth with @danasday on this. It wasn't a decision taken lightly.

This isn’t a tweet-sized problem but I’ll give it a bit of a shot: Because of the versatility of NNs, we’re seeing new possible applications every day. Because of arxiv and the availability of publishing there’s a crazy incentive for people to push existing methods forward

As a business, most software involves creating some difficult-to-reproduce IP. In the world of 2020, it’s possible to launch a new software company but only by leveraging massive investments by massive companies in the overall ecosystem

I love the tech. Have you considered a free version that adds a watermark? With an option to pay to have the watermark removed if people like the colorized image generated.

We were indeed considering that with our own app (if that happens). We're still not certain if it's a great idea though. Running these servers costs a lot of money. Encouraging lots free users to run up costs? We've seen that go bad elsewhere with DeOldify based apps.

Have u considered the ramifications of selling a service "colorize your old photos" as opposed to a technology? Service can offer lots of handy features that aren't novel tech. Free tier = watermarks, limited numbers. The raw tech can still be offered separately as non-commercial

Sure. Simple answer is that the two of us can't compete on that in terms of design/features/marketing and actually protect our IP in court effectively when the inevitable abuses start happening. And to be clear- we have absolutely no intention of growing beyond the two of us.