Prompt Engineering 101
From Zero to Fantastic Music & AI Output

Module 0 — The Universal Grammar of a Good Prompt

Every AI prompt — whether you're generating a song, a paragraph, or a spreadsheet formula — is built from the same five ingredients. Weak prompts are missing one or more of these.

The #1 beginner mistake in every tool covered in this course: vague adjectives with no anchor. "Make it epic" tells the model almost nothing — epic like a film score, epic like a stadium anthem, epic like a boss fight? Replace mood words with references + mechanics (tempo, instrumentation, structure) whenever you can.

Module 1 — Suno AI: Prompting for a Real Genre

Suno turns text into full songs (music + often vocals). It responds to two separate inputs, and beginners usually only use one of them:

• Style/genre prompt box — tells Suno how the song should sound (genre, mood, instrumentation, era, vocal type)
• Lyrics box — the actual words, which Suno also mines for phrasing/rhythm cues

1.1 — Building the Style Prompt

Think of the style box as tagging metadata, not prose. Suno responds best to comma-separated descriptor stacks layered in this order:

make an epic rock song

melodic power metal, early 2000s European scene, fast 160bpm,
soaring dual guitar leads, operatic male vocals, triumphant and
defiant, polished symphonic production

The strong version gives Suno six independent knobs to turn instead of one mood word. That's the difference between generic output and something that actually sounds like the genre you meant.

1.2 — Genre-Blending (Suno's superpower)

Suno is unusually good at hybrid genres if you name both parents explicitly rather than inventing a vague new word:

darkwave x synthwave hybrid, retro-futuristic, analog synth bass,
gated reverb drums, cold female vocals, cinematic and ominous

Avoid invented genre names with no real-world anchor ("dreampunk fusion") — Suno has nothing to pattern-match to. Two or three real genre tags blended is far more reliable than one made-up label.

1.3 — Structure Tags

Use bracketed structure/production tags inside the lyrics box to control song architecture — Suno reads these as directives:

[Intro]
[Verse 1]
[Pre-Chorus]
[Chorus]
[Verse 2]
[Chorus]
[Bridge]
[Guitar Solo]
[Final Chorus]
[Outro]

You can also drop mood/instrumentation tags mid-lyric to shift the track: [soft piano break], [drop the beat], [whispered vocals].

1.4 — Writing Lyrics That Suno Sings Well

• Keep line lengths rhythmically consistent within a section — wildly uneven syllable counts confuse the phrasing.
• Write for the ear, not the page: contractions, natural speech rhythm, and repetition (hooks) work better than literary prose.
• A strong chorus hook is usually 4–8 words, repeated — this is what Suno leans on hardest.

1.5 — Common Suno Failure Modes & Fixes

Module 2 — Claude: Precision Prompting for a Reasoning Model

Claude (and models like it) is less about "vibes" and more about giving it enough scaffolding to reason correctly. Beginners under-specify structure and over-specify tone.

2.1 — The Core Claude Prompt Shape

Write me a description for my song

Write a 2-sentence Spotify bio-style description for a synthwave
track called "Neon Exile." It's about escaping a dying city at
night. Tone: cinematic, a little melancholic. No clichés like
"sonic journey" or "takes you on a ride." Output just the two
sentences, no preamble.

2.2 — Techniques That Actually Move Output Quality

• Be explicit about what "good" looks like. Claude can't read your taste — give a positive example and a negative example when style matters.
• Ask for reasoning before the answer on anything nontrivial: "think through the structure first, then write it."
• Use tags for multi-part asks — <lyrics>, <description>, <title_options> — to cleanly separate outputs in one response.
• Iterate, don't restart. Say exactly what's wrong ("the chorus is too on-the-nose") rather than re-prompting from scratch.
• Give constraints as guardrails, not suggestions. "Keep it under 100 words" beats "try to keep it short."

2.3 — Claude for Music Workflows Specifically

Claude doesn't generate audio, but it's the best tool in this course for the scaffolding around a Suno track:

• Drafting/polishing lyrics before you paste them into Suno
• Generating a stack of style-prompt candidates for you to A/B test in Suno
• Writing release copy, track descriptions, or liner notes
• Critiquing a lyric draft against a genre's conventions

Worked example — asking Claude to generate Suno prompt candidates:

Generate 5 distinct Suno style-prompt stacks for a dark synthwave
song about a rogue AI. Each should be a single comma-separated
line, 5-7 descriptors, no duplicate genre blends across the five.
Output as a numbered list, nothing else.

Module 3 — Cross-Training: ChatGPT and Grok

Same five ingredients from Module 0 apply everywhere. What changes is personality and strengths.

Portable rule: whatever tool you're in, the failure mode is the same — vague task, no constraints, no reference point. Fix that first before blaming the model.

Module 4 — Capstone Project

Build one complete mini-release using the full pipeline:

1. Concept: One-sentence concept for a song within a chosen genre.
2. Claude: Draft lyrics + generate 3 Suno style-prompt stacks for it.
3. You: Pick/edit the best stack using the Module 1 formula; add structure tags to the lyrics.
4. Suno: Generate the track from your finished style prompt + lyrics.
5. Claude: Write a 2-sentence release description from the finished concept.
6. Review: Compare against your original genre intent using the Module 1.5 failure-mode table, revise, regenerate.

This loop — concept → Claude drafting → Suno generation → Claude critique → revise — is the actual production pipeline worth keeping as a habit, not just a course exercise.

Quick-Reference Cheat Sheet

Suno style prompt formula:

Claude prompt formula:

Universal debug question when output is bad: