you're gonna wanna record audio. 30 minutes to an hour is fine. 2 hours is the upper limit, or so ive heard. sift through it, make sure it doesn't completely suck in terms of audio quality. your singing can be garbage and it'll be fine. you can even just talk instead if you really want to. i recommend recording in japanese first, it's easy to label, which should help you get used to labeling. you should have as much variety as you can in terms of pitch and singing style. variety of pitch will expand your diffsinger's range. variety of singing styles will allow for better utilization of expressions in openutau in the future.
after recording, i typically sift through and break them up by song. its much more managable than one big recording. everything should be in wav format.