DeepL is a neural machine translation service from germany, that offers a better translation quality than traditional MTLs using Google-Translate "most of the time".
My format is basically feeding the parts that need translating to DeepL, if the result is broken, feeding them to DeepL again, if the result is still broken, feeding them to Google-Translate, and if after all that the result still remains obviously broken, translating them manually.
So more accurately "a 80% DeepL, 15% Google, and 5% Manual Frankenstein monster translation thing", but that's too long so I just call it "DMTL".
The point of it is to be able to make a "somewhat" coherent translation in a much smaller time frame than a proper translation, but without the (current) issues of fully automated translations.
(I finished over 100 DMTLs already.)