Ear training that behaves like an instrument.
Most ear trainers give you a score and a streak. SkillLab gives you a number with a unit: the smallest difference you can still hear. Not "level 12" but "you hear 2 kHz from 2.4 dB", and one line telling you what that means for your work.
STARTS EASY, DOES NOT STOP
The first question is a 15 dB difference with two answers, on pink noise. Anyone gets that right, and that is the point. From there a staircase moves with you: two right and it gets harder, one wrong and it backs off. It keeps going long past where most people arrive, down to a decibel on a third-octave band inside a full mix. There are no levels to unlock and nothing is locked in front of you.
TWELVE THINGS TO HEAR
- Louder: which of the two, down to a tenth of a loudness unit
- Which band: a real bell filter, from two choices to eleven third-octave bands
- Up or down: was that band boosted or cut
- Which filter: bell, shelf, high-pass or low-pass
- Where: balance between left and right, in decibels of difference
- How wide: the stereo image, measured as the level of the sides
- Squeezed: which one is compressed, at the same loudness
- How fast: same gain reduction, different attack time
- How much room: which one has more reverb
- How driven: saturation at matched loudness
- How long: placing an echo in milliseconds
- Which tone: naming a frequency by ear
MATCHED ON LOUDNESS, AND WE PROVE IT
Boost a band and it gets louder, so an unmatched trainer teaches you to hear volume instead of tone. Every clip here goes through the same loudness meter that LevelLab uses (ITU-R BS.1770-4) and comes out on the same target. The build fails if any pair drifts more than 0.1 LU apart. That check runs on every release.
A AND B WITH NO GAP
Both clips run at the same time, in the same loop, and the button only moves the volume. You compare on the same beat instead of waiting for a file to load, because a pause of half a second fills your memory with silence.
REAL DSP, NOT A SAMPLE LIBRARY
The filters are real biquads, the compressor really compresses, the reverb is a real Schroeder network, and the practice material is generated in the app. Nothing is a recording of an effect, so the exercises never run out and nothing is borrowed from anyone.
HONEST ABOUT WHAT IT MEASURES
A threshold here is your recognition on this phone, these headphones and this room, not a hearing test. If the number stays coarse after a few sessions, the app says so and points at HearLab instead of selling you more practice. Skill is one thing and hearing is another, and only one of those improves by trying harder.
FREE AND PRO
Every drill, every step of every ladder and your last three sessions per skill are free, without limits or ads. Nothing about the training itself is locked. Pro is a one-time purchase, no subscription, and covers what is yours rather than what the app teaches: the whole history with the curve behind it, your own music as a practice source, and a PDF report with your name on it. Elsewhere this kind of training costs a subscription per year.
NO MICROPHONE
SkillLab never listens. There is no microphone permission, no account, no analytics and no advertising code in the app.
THE AUDIOLAB FAMILY
Each app measures one honest thing, on your device, with nothing uploaded:
- HearLab: your ears, left against right
- MicLab: your microphone, checked in 12 seconds
- RoomLab: your room's acoustics, from a clap
- LevelLab: loudness the way streaming platforms hear it
- NoiseLab: how loud it is where you stand
- SignalLab: every test signal, tones to sweeps
- TuneLab: a tuner with cents as a number
- SkillLab: the ears all of it passes through
Ear training on real filters. Ends in a threshold in dB, not a score.