Home Chinland InfoTech Academy
Ready
All Courses

Drop the Beat

How AI turns words into music — and text into a human voice. Pick any concept. Many of these play real sound — click “enable sound” up top first. Press Esc anytime for this menu.

The foundation

Sound is a wave

Every sound — a note, a voice, a drum — is just air wobbling fast. Draw that wobble and you get a wave. Faster wobble = higher pitch; bigger wobble = louder. Enable sound, then drag the sliders to hear and see it.

Pitch (frequency): 220 Hz

How fast the wave wobbles → how high it sounds.

Volume (amplitude): 50%

How tall the wave is → how loud it is.

Shape

The wave's shape = its “timbre” (flute vs buzzer).

Stacked stems

A song is layered tracks

A song isn't one sound — it's layers stacked together: drums, bass, chords, melody. Toggle each layer and hear the track build. AI music tools generate these “stems” and mix them, just like a producer.

Build the track

🥁 Drums
🎸 Bass
🎹 Chords
🎵 Melody

Same trick — meet AI

Each layera “stem”
Stacked togetherthe full song
AI generates eachthen mixes them
Diffusion, for audio

From noise to music

Remember how image AI turns static into a picture? Music AI does the same with sound. It starts with a hiss of audio noise and, guided by your prompt (“lo-fi, rainy, chill”), cleans it into a musical phrase — step by step. Drag the slider.

Denoising step: 0%

Left = pure hiss. Right = clean musical phrase.

Prompt

“lo-fi hip-hop, rainy, chill”

The words steer what the noise becomes — same as images.

Same trick — meet AI

Audio hissrandom noise
Clean it updenoising
Your wordsthe steering
Like next-word

Predicting the next note

One way AI makes melodies: like it predicts the next word, it predicts the next note. Give it a few notes and it continues the tune in a style it learned. Play a seed, then let AI carry on.

Seed melody

Same trick — meet AI

A few notesthe “prompt”
Most likely next notethe prediction
Repeata whole melody
The TTS pipeline

Text becomes speech

Type a sentence and AI turns it into a talking voice. Under the hood: it splits your words into little sounds (phonemes), predicts the melody and rhythm of speech, then generates the audio wave. Enable sound and press speak.

Your text

The pipeline

Your wordstext
Split to soundsphonemes
Add rhythm/melodyprosody
Generate waveaudio out

Uses your device's built-in voice for the demo. Real tools (ElevenLabs etc.) sound far more natural.

Then vs now

Why it sounds human now

Old text-to-speech was flat and robotic — every word the same length and pitch. Modern AI adds prosody: stress, rhythm, pitch changes, even breaths and emotion. Drag through the years to see (and hear) the leap.

Year: 2010

2010 — flat, robotic, monotone.

What changed

Newer models predict the natural ups and downs of real speech — so a sentence rises, falls, pauses and breathes like a person.

A voice fingerprint

Cloning a voice

Give AI a short sample of someone talking and it learns a “voice print” — the unique fingerprint of that voice. Then it can make that voice say anything, including words they never recorded. Powerful — and the reason consent matters (next section).

Step through it

1 · A short sample

Just a few seconds of someone speaking is enough for modern tools.

Same trick — meet AI

Voice samplethe input
Unique featuresthe “voice print”
+ any new textthat voice, new words
Steerable delivery

Emotion & style dials

The same sentence can be said a hundred ways. AI voices are steerable: you set the emotion, speed and pitch. Pick a delivery and hear the difference — the words stay the same, the feeling changes.

Delivery

Speed: normal

Pitch: normal

Same words: “I can't believe you're here right now.”

The big one

Consent & voice deepfakes

If AI can clone a voice from seconds of audio, it can make someone “say” things they never said — for scams, fake news, or impersonating family. The skill isn't just making voices; it's using them responsibly, and spotting fakes.

Scenario

📞 Scam call

A cloned voice of a “relative” calls in a panic asking for money. Always verify another way.

The rules

✅ Get consent before cloning a real voice. ✅ Disclose AI voices. ❌ Never impersonate to deceive. 🛡️ Agree a “safe word” with family for phone scams.

Rights & money

Who owns an AI song?

You typed the prompt — but who owns the result? It depends on the tool's terms, your plan (free vs paid), and what you asked for. “In the style of [real artist]” is a legal grey zone. Here's the lay of the land.

Check

Free tier

On most tools, free-tier songs are for personal use only — the platform may keep rights, and you often can't monetize. Read the terms.

Sets up Day 5

If you want to earn from AI music, this matters. We'll cover monetizing properly on Day 5.

Stereo & panning

Why headphones feel 3D

Two ears, two speakers. Send a bit more sound to one side and your brain places it there in space. That's “panning.” Put headphones on, enable sound, and drag the sound left and right.

Pan: centre

Slide left/right — the sound moves between your ears.

Same trick — meet AI

More to one sideleft/right balance
Your brainplaces it in space
Many sounds panneda wide mix
The actual skill

Prompting a music tool

Tools like Suno don't need sheet music — they need a good description. A vague prompt gives a thin, generic loop; stack genre + mood + instruments + tempo and the track fills out. Toggle the parts and press play to hear the difference.

Build the prompt

🥁 Genre — adds the beat
🌧️ Mood — adds atmosphere
🎹 Instruments — adds melody
🐢 Tempo — sets the speed

Richness: thin & generic

BPM

Beat & tempo

Tempo is the heartbeat of a track, measured in beats per minute (BPM). Slow = chill; fast = energetic. Same drum loop, different speed — drag the BPM and hear (and watch) the pulse change.

Tempo: 90 BPM

Mid-tempo — a steady groove.

Rough guide

60–80 BPMlo-fi, ballad
90–110 BPMhip-hop, pop
120–130 BPMhouse, dance
140+ BPMdrum & bass
Styles sit close

A map of genres

Just like words and pictures, AI puts musical styles on a “map” — similar genres sit near each other. That's how it blends “lo-fi + jazz” or slides from pop toward rock. Click a genre to see its neighbours.

Pick a genre

Lo-fi

Sits near jazz and chillhop — mellow, mellow neighbours. Blend two nearby points and you get a fusion.

Same trick — meet AI

Each genrea point
Similar stylessit close
Between twoa blend
AI that sings

Lyrics become vocals

The headline trick of tools like Suno: give it words, and it sings them — matching each syllable to a note, over the music. Watch the lyrics light up in time with a melody, syllable by syllable.

Lyrics

“Ci-ty lights are call-ing me to-night”

Same trick — meet AI

Each syllablemapped to a note
Note + voicea sung line
Over the beata full song

Demo uses your device voice as a stand-in singer; real tools sing in tune.

Stems & remix

Pulling a song apart

The reverse of layering: AI can take one finished song and separate it back into its parts — vocals, drums, bass, other. That's how remixes, karaoke tracks and mashups get made. Click to split, then solo a part.

Step

Same trick — meet AI

One mixed songthe input
AI separatesthe stems
Mute/keep partsremix / karaoke
Foley & SFX

Sound effects from words

Not just music — describe a sound and AI makes it: “footsteps on gravel,” “thunderclap,” “sci-fi door.” Perfect for the videos you made on Day 2. Pick one and hear a quick synthesized version.

Describe a sound

Why it's useful

Add atmosphere and effects to your AI videos without recording anything — just describe what you need.

Menu ← → move · Esc menu · 1–9 jump