Poster Comment:

The tech is basically 'deepfake on the fly'. It allows you to talk while the AI translates into different languages by modulating a form of your voice after it has "learned" and synthesized it. Then it also superimposes a sort of deepfake mouth to match the spoken translations.