OctoMonic
OCTOMONIC
28 Jul 2026

Aminadav Street 9

Vocal mixed over Gotcha — RunGood

Mix session 01 · v3 — rebuilt vocal chain · timing locked at −0.598s · 147.66 BPM · −14.0 LUFS

Timing is unchanged — you confirmed it's right, so it stayed at −0.598s. Everything below is the vocal chain rebuilt from scratch. Two directions: pick one and I'll refine it.
A  ·  PRODUCED

More space

Fuller reverb and delay, vocal sits into the track.

Open the file directly ↗
B  ·  UPFRONT

Drier and closer

Much less reverb, more presence, vocal sits in front.

Open the file directly ↗
If the players don't play on your phone:
  1. Check the ring/silent switch on the side of your iPhone. When it's set to silent, Safari mutes inline audio on a web page but gives no warning — it just looks broken. This is the most common cause.
  2. Tap “Open the file directly” above. That hands the MP3 to your phone's own audio player instead of the in-page one, which also lets you save or share it.
  3. Turn the volume up with the physical buttons while audio is playing — iOS keeps a separate media volume that can sit at zero.

What was wrong, and what changed

WasProblemNow
Haas double Delayed copies of a mono voice at 14/23 ms panned L/R — comb-filters and smears transients. Real doubles are separate takes. Removed. Width comes from stereo reverb and delay only.
Exciter + soft-clip Stacked harmonic distortion on a 63 kbps source amplifies codec artifacts and sibilance into harshness. Both removed.
Noise-burst "plate" Decayed white noise isn't a reverb — grainy and metallic. Measured 36% tail density where diffuse is ~85%. 16-line feedback delay network with 8-stage diffusion. 80% density, RT60 1.43s.
FFT denoise on the voice Leaves watery artifacts, and at −57 dB the noise floor was already masked by the beat. Gone from the dry path. A gentle gate runs only on the sends.
—Vocal lacked density and consistency. Parallel compression — a heavily compressed copy blended underneath.
—Reverb and delay cluttered the words. Ducked sends — sidechained to the dry vocal, so they bloom in the gaps and stay out of the way during lines.
Beat ducking too hardAudible pumping. Ratio 4→2.5, release 200→320 ms.

Measured, not asserted

Metricv1 (old)v3 producedOriginal beat
2–6 kHz harsh-band energy4.14% 3.23%3.28%
Mono-sum loss−0.25 dB −0.18 dB−0.43 dB
Crest factor (dynamics)15.08 dB 15.23 dB15.95 dB

The harshness fix is the real one — v1 was pushing nearly a full percentage point of extra energy into the fatiguing band, and v3 now matches the beat itself. The mono-compatibility gain was small; removing the Haas double was still correct, but it wasn't the main offender.

Reference

v1 — the version you said sounded bad
Kept for comparison so the change is audible.
Open the file directly ↗
Your combined recording — the timing source
Beat at 3.6612s, vocal at 3.0635s.
Open the file directly ↗
Original beat — untouched
−13.6 LUFS, true peak +0.1 dBFS, loudness range 2.7 LU.
Open the file directly ↗
Your raw vocal — untouched
−27.2 LUFS, peak −8.9 dBFS. Clean take, 63 kbps mono, hard 16 kHz codec ceiling.
Open the file directly ↗

What I still can't fix from here