Your course audio is clear and you follow it easily. Then a real person says the same thing and it arrives as one long sound. They are not speaking faster than the recording — not by much. They are speaking differently.
Words stop having edges
Course audio pronounces each word as if it were alone. Real speech runs them together: the end of one word borrows the start of the next, unstressed vowels shrink to almost nothing, and whole syllables drop out. So the problem is rarely speed. It is that the shapes you memorised — the careful, isolated versions — never appear in the wild.
Train on the messy version deliberately
Listen to the same thirty seconds of natural speech four or five times. First for the gist, then for the sentence boundaries, then for the words you missed. The fourth pass usually snaps into place. Then read the transcript and listen again. The gap between what you heard and what was actually said is the exact thing you are learning.
Say it the fast way yourself
Producing the reduced forms teaches your ear to expect them. Imitate a short phrase at full speed, running the words together the way the speaker did, rather than pronouncing each one correctly and separately. This feels sloppy and wrong at first. It is neither. It is what the language actually sounds like.
You do not need to get faster at decoding. You need to learn the shapes that real speech uses, and the only way in is repeated contact with the unedited version.