Defoko was made with samples from another vocal synthesizer, AquesTalk, which if I recall correctly has the voice synthesized from the ground up without any samples to begin with.
So if you want authenticity, you will have to find a sample-less vocal synthesizer, or start researching the source-filter voice synthesis model to make your own.
...
it would be funny if you actually did
Reduce your vocal quality in a program such as Tenacity (or an older fork of Audacity without the suswarez) by reducing the project rate and re-rendering (make sure to hold on to the original files), then try introducing some noise
Playing around, I found the Expand and Compress distortion type in Tenacity
anyways I'm not an audio engineer but this should bring you closer
ok break a leg bye