Phase 4: Audio & Music Generation · 50 min · Python · Demucs · pydub

Audio Generation Engineering

Stem separation, structure control, voice cloning — the engineering layer that turns raw generation into production audio.

Hiring signal: Audio engineering skills (stem separation, structure control, format handling) demonstrate you can work with audio as an engineer, not just generate it.

What you will learn

Differentiate text-to-audio (SFX) vs text-to-music (full songs)
Handle audio formats: WAV, MP3, FLAC, sample rates (44.1kHz, 48kHz), bit depth
Implement stem separation: splitting generated audio into individual instrument tracks (Demucs, Moises)
Control music structure: intro/verse/chorus/outro, mood prompts, tempo control

The Problem

A team generates a song with Suno. The song sounds good, but:

They need to remove the vocals and keep only the instrumental — they don't know about stem separation
They need the audio in 48kHz WAV for a video project, but Suno outputs 44.1kHz MP3 — they don't know how to convert
They need the song to have a specific structure (intro → verse → chorus) — they don't know how to control structure
They need to clone a specific voice for the vocals — they don't know voice cloning

Audio engineering skills (stem separation, structure control, format handling) demonstrate you can work with audio as an engineer, not just generate it.

What you'll build
Implement stem separation with Demucs to split generated audio into individual instrument tracks. Handle audio format conversion (WAV, MP3, FLAC, sample rates). Control music structure with prompt engineering. Document audio engineering best practices.

Audio Formats

Common Formats

Format	Use Case	Quality	Size
WAV	Master/archive	Lossless	Large
FLAC	Archive/storage	Lossless compressed	Medium
MP3	Delivery/streaming	Lossy	Small
Opus	Streaming/real-time	Lossy, efficient	Smallest
AAC	Streaming (Apple)	Lossy	Small

Sample Rates

Rate	Use Case
44.1 kHz	CD quality, music standard
48 kHz	Video/film standard
22.05 kHz	Voice-only (low quality)
96 kHz	High-resolution audio

Bit Depth

Depth	Dynamic Range	Use Case
16-bit	96 dB	CD quality, standard delivery
24-bit	144 dB	Production/mastering
32-bit float	Unlimited	DAW, processing

Format Conversion with pydub

from pydub import AudioSegment

# Load MP3
audio = AudioSegment.from_mp3("generated_song.mp3")

# Convert to WAV at 48kHz, 24-bit
audio = audio.set_frame_rate(48000)
audio = audio.set_sample_width(3)  # 24-bit = 3 bytes
audio.export("output_48k_24bit.wav", format="wav")

# Convert to Opus for streaming
audio.export("output.opus", format="opus", bitrate="128k")

You generated a song with Suno (44.1kHz MP3) but need it in 48kHz WAV for a video project. How do you convert it?

Use pydub: audio = AudioSegment.from_mp3("song.mp3"); audio.set_frame_rate(48000); audio.export("output.wav", format="wav"). This converts the sample rate from 44.1kHz to 48kHz and the format from MP3 to WAV. You can't just rename the extension — the actual audio data needs to be re-encoded.

Unlock the full lesson

You've read the first 2 sections. The rest of this lesson covers Stem Separation, Music Structure Control, Voice Cloning for Music, Audio Analysis with librosa, Key Takeaways, What's Next — plus a hands-on lab, quiz, and project artifact.

Create a free account to unlock Phase 0 and Phase 1 of every course — no credit card.