Slow a track down to work it out by ear

It opens at 80%. That makes the track 1.25 times longer, and a 120 BPM song plays at 96 BPM. The key slider stays at ±0 on purpose: working something out by ear means naming the pitches you hear, and moving the key makes every answer wrong by the same amount.

Slower is not steadily clearer. The tool works by splitting the audio into short pieces and spacing them further apart, so the further you stretch, the more the seams show — drum hits and pick attacks smear. Those attacks are exactly what you listen to for rhythm. Somewhere around 70–80% is the line between "slow enough to follow" and "smeared enough to guess", and at 50% the pitches survive but the articulation does not.

You rarely need the whole song slowed. Trimming the few bars you are stuck on before bringing them here makes the render faster and the loop easier to live with.

QHow is this different from 0.75x on a video site?

For listening alone, the player is enough. Two things differ here. The slowed version comes out **as a file**, which matters when your hands are on an instrument or when you want to loop a section in another app. And the steps are finer: players offer 0.75 and 0.5, while this takes any whole percent, so you can sit exactly on the line where 85% is followable and 80% is not.

QCan I slow down just the bass, or just the drums?

Not on this page alone, but in two steps: separate the track with the stem splitter first, then bring the part you want here. For transcription that pairing is the single most effective thing you can do. Slowing a full mix stretches every part at once and the smeared results overlap; a single part keeps its edges much longer.

QIs there any way to avoid the artefacts?

Not entirely — the method is to cut the audio into short pieces and space them out, and the spacing is where the seams come from. You can reduce it, though: stay at or above 80%, and separate to a single part before stretching. The worst combination is a dense full mix taken all the way down to 50%.