One idea suggested was a translation caching service. Unsure how the economics work, however the same translations are being requested over and over to DeepL at present.

Maybe a layer in front with a small fee could offset costs or KYC issues.