Three music-production exercises.

01 / About 5 minutes

Hear straight and dotted delay at 72 BPM.

These original synthetic pluck demos use the same notes and source level. Each processed example adds one quieter repeat. There is no feedback loop, so the repeat spacing is the main difference to listen for.

1. Listen to the dry phrase, then each delayed version at a comfortable level. Tap a steady quarter-note pulse.
2. Open the delay and reverb calculator, enter 72 BPM and compare Straight with Dotted.
3. If you have a DAW, import the dry WAV and set a delay to one of the values below. Keep feedback at zero and the first repeat quieter than the source.

Check: A quarter note is 60,000 ÷ BPM milliseconds. A dotted eighth is three quarters of that duration. Which repeat fills more of the gap after each note? Which would leave more room for a second part?

Go deeper: delay timing, feedback and stereo placement.

02 / About 5 minutes

Build a beat, then remove one part.

1. Open the online beat maker. Choose Blank canvas, or Clear the pattern, and set the tempo to 120 BPM.
2. Across its sixteen-step bar, put kicks on steps 1, 5, 9 and 13. Put snares on 5 and 13, then closed hats on 3, 7, 11 and 15.
3. Play the bar. Switch the hat steps off, listen again, then restore them. Keep the kick and snare unchanged.
4. Use the BPM tapper while listening. Compare your taps with 120 BPM. If you get about 60 or 240, check which pulse you are following.

Check: The kick anchors every quarter note; the hats sit between kicks. Can you describe what disappears when the hats stop? A useful answer names the rhythmic role, rather than just saying one version is better.

For a collaborator, export MIDI and load a drum instrument in the receiving DAW. MIDI stores note instructions, so it will not sound by itself. Read the four-on-the-floor guide for the next step.

03 / About 10 minutes, plus processing

Check a separated stem against its source.

Use a short vocal-and-instrument recording you own. Ideally, make a mix from your own vocal and backing tracks and keep those originals for comparison. This gives you a known reference instead of treating the separator’s output as ground truth.

1. In the AI stem splitter, choose Vocals only, select your short mix and create the selected stem.
2. Compare the separated result with the original mix at a similar listening level. Listen near consonants, cymbal hits and quiet gaps.
3. If you kept the original vocal track, compare that too. Note one place where the estimate preserves the voice and one place where you hear bleed, missing detail or a watery texture.
4. Write down the clip length, browser, device and observed processing time. If the device struggles, use a shorter clip or move to a computer.

What this demonstrates: separation estimates parts from a mix. It does not recover the original recording session. Keep that distinction clear when sharing files with a teacher or collaborator. See stems versus multitracks.

Your listening notes.

Name / date:

One setting I changed, and what I heard:

One result I would check with a collaborator:

For teachers and collaborators

Ask learners to explain their comparison before showing the expected answer. Keep the same source and playback level when comparing effects. For younger learners or a short class, start with the delay and drum exercises; leave audio separation for a supervised session on a suitable device.

Share this page or print it for your lesson. The three WAV files are synthetic demonstrations created for these exercises, not recordings of a Kulture event or claims about an artist’s workflow.

Reference: projectkulture.com/lab/practice/