Or rather, “Don’t, unless there’s a very good reason, and that reason is not your belief that using authentic audio will somehow get beginners to understand authentic audio.”
Practice with fast, accented audio is NOT the most effective way to help people under stand fast, accented speakers, with the exception of that particular audio clip.
The biggest predictor of listening comprehension — including accented or fast speech — is vocabulary size. That means acquired, not memorized, vocabulary. Vocabulary that is automatically transformed into meaning.
I train interpreters. We are very invested in listening comprehension, and for a wide range of very authenic speakers. I have yet to see research that shows that “practice” on things that cannot be understood is an effective means of getting people to understand those things, except insofar as they might be practicing guessing skills. (That is actually something we teach in interrpreting — but please note that we also have to teach “un-guessing” skills, or how to backtrack during your interpretation and say, no, that’s not what I “meant” due to a wrong guess.)
I truly cannot see how the advantages to be had during, say, 1-3rd years, from “authentic” audio outweigh those coming from extensive input in terms of actually building listening comprehension. This is because the barriers to understanding rapid or accented speech feed into the Effort Model (this is for interpreters, but it holds just as much for plain old listening without atlking at the same time). Giles’ Effort Model basically holds that your brain is only so big. If it takes all your brain attention to decode learner-level speech, inputting harder speech (faster, accented, unknown vocabulary) will cause a breakdown due to insufficient processing resources.
The way to free up more resources in the brain is to make as much automatic as possible. Acquiring vocabulary makes its transition to meaning automatic, freeing up resources to handle odd challenges such as accent or speed. AND vocabulary knowledge benefits other areas of language as well. Training on “samples” of native speech do not.
The road to good listening comprehension — if you want to move forward, instead of simply sitting in a sportscar for photographs (which may be necessary, depending on your situation) is still comprehensible input, not training and practice. You can tell a first-year student “Taiwanese in the south often substitute “f” for “h”” but it won’t help until they have the language automatically enough to have the resources to APPLY THE RULE. I have been speaking Chinese for 30 years and know this rule, and can still “trip” when it pops up unexpectedly.