Translation filters, hearing the kick on a phone before you bounce

Three years ago I finished a track. Mastered it on Genelec 8050s, A/B’d it through my Apollo against three reference tracks, signed it off, sent it to the label. The label put it on Beatport. Three days later I listened to it on my phone in a taxi and the kick was gone.

Not muted. Gone. The fundamental that I’d spent forty minutes dialling in on the studio monitors was inaudible on a phone speaker. Worse, the mid-bass thump that filled in for the missing fundamental was sloppy, it sounded like the kick was over-compressed. It wasn’t over-compressed. The mid-bass thump was the only part the phone could reproduce, and I hadn’t voiced it.

The track had been released. There was no recall.

That afternoon, I wrote down five questions:

1. Why didn’t I check the kick on a phone?
2. How could I have checked the kick on a phone faster?
3. What would have made me do it?
4. What’s the right number of “translation systems” to check?
5. Could the plugin do this for me?

Three years later, MrKicksy ships with translation filters. Here’s what we built, why each one, and what we still can’t do.

What a translation filter is

It’s a filter chain that emulates a specific playback system. Phone-speaker emulation is the most aggressive: a high-pass at about 200 Hz, a notch at about 3 kHz, a low-pass at about 6 kHz, and a soft saturator to simulate the speaker’s nonlinearity. You feed it your full-range master. It outputs what a phone would reproduce.

The point isn’t for you to listen to your music on the filter chain as a daily monitor. The point is to flip to it for thirty seconds before you commit, and hear what a large share of your listeners are going to hear.

The five we ship

We argued about which systems to emulate for a month. Here’s where we landed.

Club PA. A loud, low-end-extended system with sub. Most kicks sound great on this, it’s what we mix for in the studio. We include it not because it’s the hard case but because it’s the reference. You want to hear “this is what it sounds like at the venue I built it for”, then compare.

Studio monitors at low volume. This is the one most producers forget. Your monitors sound great at 75 dB SPL, that’s where you mix. They sound different at 60 dB, which is much closer to how people listen at home. The Fletcher-Munson curve flattens the low end at low volume. Bass-light mixes sound thin at conversational level. The filter emulates this by attenuating below 80 Hz proportionally to perceived loudness. If your kick disappears, you’ve over-relied on the sub.

Headphones, consumer. Cheap consumer headphones (the kind that come bundled with phones, or the £20 set somebody bought because their good ones broke on tour). Bass-heavy, mid-scooped, harsh in the 5 kHz region. Most “the kick is too loud” complaints come from people listening on these.

Phone speaker. The hard case. Almost no sub, almost no low-mid, mid-forward, harsh transients. If your kick translates here, it translates anywhere.

Car system. A different hard case. Cars have boom around 60–80 Hz from cabin resonance, mid-bass dropout around 150–250 Hz, presence boost around 3 kHz. Kicks that sound “clean” on monitors often sound “muddy” in cars because the cabin adds bass the producer didn’t account for.

What we don’t emulate

A lot of things. Specifically:

We don’t emulate streaming codec damage. The MP3 / AAC encoder will do its own thing to your kick, especially in the high end. We considered including it; we decided against it because the encoding ladder differs per platform (Spotify, Apple Music, Beatport WAV stay full-range; SoundCloud and YouTube Music compress differently) and modeling every codec would require a small DSP team. We point you at the platform-specific reference renders the streaming services publish.

We don’t emulate the listening environment. A phone speaker in a quiet room sounds different from a phone speaker in a coffee shop. We can’t model the noise floor of every place your listener might be. The phone-speaker filter assumes worst-case (the coffee shop) because that’s what you should be designing for.

We don’t emulate the human ear. That sounds glib, but it’s real. Hearing is itself a filter. People who’ve been to a lot of loud shows have their own personal frequency response curves. The filter chain models the speaker, not the listener.

What this is for

Five filters, thirty seconds each, before you commit. If your kick survives all five, the release won’t embarrass you on a phone in a taxi. If it doesn’t survive one of them, you tweak before the bounce, not after.

That’s the entire pitch. There’s no clever DSP. The filters are textbook. What we shipped is the workflow: a button you press to hear what your listeners will hear, fast enough that you press it.

We’d have shipped this regardless of the synth design. But it pairs naturally with a kick instrument: when the whole tool is about the kick, you should be checking the kick on every system that matters.

The track three years ago is still on Beatport. It still sounds bad on a phone. I learned from it. So, I hope, will you.