1/10

The studio desk at night: an SM7B mic, an SSL2+ interface and an EQ under the monitor.
The room from the homepage, with the audio interface on the desk.

07 · Audio engineering · how I learned to mix

Dec 2019. Terrible covers. Terrible presets.

The story of how I learned to mix, from that first cover in 2019 to the chain on my desk today.

AudioMaking music since Dec 2019

The only way to learn: do it.

  1. My own vocals
  2. Friends’ vocals
  3. Friends’ sessions

real material to practice on, from day one

Every preset, rebuilt from scratch.

from zero

EQ, compression, saturation, gain (the basic tools of a mix), every knob learned by turning it

Less is more.

every plugin

one move

mixing for the record, not for me

Study the greats. Then figure out how.

Serban Ghenea and Jaycen Joshua: studied by ear, never worked with

Years training my ears.

2019

Now

illustrative: the same idea, mixed then and now

Mixing first. Mastering and producing follow.

  1. Mixing
  2. Recording
  3. Producing
  4. mastering comes with the mix

where the hours go, my own estimate: mixing takes 2× to 4× the time of recording or producing

The chain on my desk.

The desk in the room: the SM7B, the SSL2+ interface and the EQ.
The same desk from the room, behind the signal chain.
  1. SM7Bthe mic
  2. SSL2+the interface
  3. EQthe EQ
  4. HS8sthe speakers

the same chain shaped dub’s mixer and OTD’s never-touch-the-audio rule

Now: guitar and piano.

guitar classes, and a MIDI keyboard that’s teaching me piano

Train your ears on one section.

Silent: drawn from the notes and drum hits in OTD’s own analysis of “What If? (Pt. 2)” (Eli the Kidd). Nothing plays, and no audio file ships with the site.

CH 1 · EQ / DYNWhat If? (Pt. 2) · analysis, not audio

Turn a knob to reveal the decision.

44 log-spaced bands, 32 Hz–16 kHz, drawn from the real notes and drum hits in OTD's own analysis of this section (no audio file ships with the site). The three knobs are illustrative mix moves for a section shaped like this one, not a leaked session file. Keys on the fader: ← → scrub.

The only way to prove a mixer adds nothing.

CODE 7.1dub/tests/test_render.py:37-46
def test_render_without_takes_reproduces_the_original_audio(tmp_path):
    client = TestClient(app)
    r = client.post("/api/render", data={"pack_id": "diner-argument"})
    assert r.status_code == 200, r.text
    (tmp_path / "out.mp4").write_bytes(r.content)

    to_wav(tmp_path / "out.mp4", tmp_path / "out.wav")
    to_wav(FIXTURE / "video.mp4", tmp_path / "orig.wav")
    peak, _ = peak_rms_db(str(tmp_path / "out.wav"), str(tmp_path / "orig.wav"))
    assert peak < -90, f"render colored the audio: residual peak {peak} dBFS"

With no takes, the mix is the original audio. It has to come back identical, not merely close.

Counted, with the date they were counted.

null-test residual, dub’s zero-take render
< −90 dBFSas of 22 Sep 2026, describes dub’s mixer
per-take gain clamp, one static gain, never a compressor
±15 dBas of 22 Sep 2026, describes dub’s mixer
per-take timing shift, only when it clearly helps
±300 msas of 22 Sep 2026, describes dub’s mixer

What I don’t claim.

What’s on the desk.

Proves freelance mixing and mastering, the audio discipline carried into two other items (résumé → Freelance Audio Engineer).

Elsewhere in this portfolio

  1. 01 Résumé
  2. 02 Bishop
  3. 03 OTD
  4. 04 dub
  5. 05 Percival
  6. 06 Team game
  7. 07 Audio
  8. 08 Teaching
  9. 09 Sacred Lands
  10. 10 Skills
  11. 11 About