Making a 47-Minute Mix Survive Phone Speakers

A friend asked me to master a 47-minute mix for a small tape run: 11 tracks, 44.1 kHz/24-bit WAV out of a borrowed interface, peaks at -0.3 dBFS on the bus, and it sounded fine on my monitors. On a phone speaker, the sub-bass lines vanished and two transitions collapsed into a single low rumble. That was the whole problem, in one sentence.

I did not remaster. I built a monitoring check first. The chain was ffmpeg for the band split, sox for level measurement, and a fixed set of three playback profiles I could switch between quickly:

ffmpeg -i mix.wav -af "highpass=f=120,lowpass=f=6000" phone_ref.wav
ffmpeg -i mix.wav -af "highpass=f=60,lowpass=f=12000" laptop_ref.wav
sox phone_ref.wav -n stat 2>&1 | grep -E "RMS|Maximum amplitude"

The 120 Hz highpass is aggressive on purpose. If a transition depends on content below 120 Hz, the phone ref will show it as a gap. I ran all 11 tracks plus the transitions through it, marked the timestamps where the RMS dropped more than 6 dB relative to the previous 10 seconds, and got four spots. Two were real problems, two were quiet intros that were fine.

For the two real spots, I added a narrow midrange presence band on the transition only, not the whole mix:

ffmpeg -i mix.wav -af "atrim=start=1120:end=1180,asetpts=PTS-STARTPTS,\
 equalizer=f=800:t=q:w=1.2:g=3.5" trans_fix.wav

Then I spliced it back with a 40 ms crossfade. Three and a half dB at 800 Hz is enough to keep the kick audible on a phone without making the mix sound like a podcast on real speakers. That is the trade: you are buying phone intelligibility with a small amount of midrange congestion.

Where it breaks: this method lies on material with intentional sub-only drops, because the highpass check will flag them as dead spots. I also tried the same approach on a 6-minute ambient piece and stopped, since the 800 Hz boost turned pad tails into a hiss I could not unhear. And the RMS threshold is arbitrary; 6 dB worked here because the mix was already fairly consistent, but on a dynamic live recording I would have to look at the actual waveform instead.

What I still do not know is whether the tape will be played on the same devices I tested. Does anyone here have a cheap reference setup for phone-speaker checks that does not involve actually owning five phones?