Mix for phone speakers
AI-generated songs are often mastered with the vocal only about 1 dB under the backing track. Phone speakers boost the 2–5 kHz range where the voice lives, so on a phone the vocal sounds harsh and the music disappears.
Steps
- Split stems. Separate the song into vocal and instrumental.
- Vocal. Lower 3–4 dB. Cut 2–2.5 dB around 2.8 kHz and about 1.5 dB around 5.5 kHz. High-pass at 100–120 Hz.
- Instrumental. High shelf +2 dB above 8–9 kHz for air, +1 dB around 3.5 kHz for detail, −1 dB around 250 Hz to remove mud.
- Bus. Gentle compression (2:1), then normalise to about −14 LUFS integrated with true peak at −1 dBTP or lower (the level YouTube and Instagram play back at).
- Listen on a phone. Speaker, not headphones. The words should be clear but the drums and synths should still be easy to follow.
ffmpeg sketch
[vocal]volume=-3.5dB,equalizer=f=2800:t=q:w=1:g=-2.5,equalizer=f=5500:t=q:w=2:g=-1.5,highpass=f=110[v];
[inst]equalizer=f=9000:t=h:w=0.7:g=2,equalizer=f=3500:t=q:w=1.2:g=1,equalizer=f=250:t=q:w=1:g=-1[i];
[i][v]amix=inputs=2:normalize=0,acompressor=threshold=-14dB:ratio=2,loudnorm=I=-14:TP=-1.2