🎛️ Machines That Hear · Day 16

Tiny Audio, Big Tradeoffs

To send your voice over the internet, computers have to shrink it into a small file. Shrink a little and it still sounds great; shrink too much and it turns into a robotic, muffled mess. Let's crush some audio and watch quality fight file size.

↓ crush it
🏷️ The pro wordsHow many times per second we measure the sound is the sample rate (in Hz/kHz). How finely we measure each point is the bit depth. Squeezing it into a small file is a codec (compression), and the data-per-second is the bitrate. Fun fact: to capture a sound you must sample at least twice its frequency — the Nyquist rule!
🎛️ Lab · The Audio Crusher

Quality vs file size

🟩 is the original sound. Lower the sample rate and it gets blocky (high notes vanish). Lower the bit depth and it gets steppy and noisy. Both make the file smaller — the eternal trade-off engineers balance!

44 kHz

sample rate — lower = blocky, loses high notes

16

bit depth — lower = steppy, adds noise

File size
100%
Quality
100%
🎯 Challenge: shrink the file below 40% while keeping quality above 70% — that's the "just right" a codec aims for!
🚀 Beyond the Basics

How the pros really do it

Crushing the sample rate and bit depth shows the trade-off — but it sounds ugly. Real codecs are far cleverer:

🟢

Crush it (what you did)

✅ Easy to understand. ⚠️ Sounds harsh — you throw away sound you actually wanted.

🎧

Smart codecs

Use psychoacoustics — only throw away sounds your ears can't hear anyway. Opus (calls/Discord), MP3, AAC. ✅ Tiny files, still sound great. ⚠️ Complex math.

🧠

Neural codecs (AI)

AI shrinks voice to super-tiny sizes with surprising quality — great for slow connections. ✅ Amazing compression. ⚠️ Needs a model.

🏷️ Pro names to look upOpus (the modern voice-call codec) · MP3 / AAC · EnCodec & SoundStream (neural audio codecs).

🎉 Day 16 Complete!

You learned how digital audio works — sample rate, bit depth, and codecs — and the constant trade-off between sounding good and staying small.

Day 17: Listen This Way →
Tomorrow: smart speakers use several mics to hear you across the room. Meet beamforming. 🎯
AI Adventure · Machines That Hear · Day 16 — made for young inventors 🚀