I feel like they failed to do proper incentive modeling for crowdsourcing (or maybe I just don't understand the target audience).

Recaptcha(v1) also does transcription correction but they incentivize a huge pool of users to help, not because they want to help, but because they need to prove they aren't robots.

Here we need to rely on volunteers? I am not sure if transcribing old documents is rewarding enough for people to do it out of good will. It's not low effort like Folding@home where you can just set and forget. It requires a lot of repetitive and lonely effort!