How to Use Suno to Write a Custom Song From a Prompt
Most people open Suno for the first time, type something like "happy pop song," listen to the result, and then close the tab thinking the tool is a novelty. That's a waste of what's actually a capable music generation system. The difference between a throwaway result and something you'd actually use in a video, podcast, or product demo is mostly in how you write the prompt and how you structure the lyrics.
Suno generates both audio and lyrics simultaneously from a text description. You can let it write everything, or you can bring your own words and just ask it to handle the music. Both approaches work, but they require different prompting strategies.
Style Prompts: How to Describe the Sound You Want
The style prompt is not a song title. It's a description of the sonic feel. Think about genre, instrumentation, tempo, mood, and era. The more specific you are, the more useful the output.
Vague prompt: "upbeat pop song about summer"
Better prompt: "indie pop, female vocal, acoustic guitar, tambourine, breezy, 2000s feel, like early Katy Perry without the production excess"
Notice the second prompt includes a reference point. Suno doesn't reproduce copyrighted songs, but it understands stylistic references and uses them to calibrate tone, tempo, and arrangement choices. You can stack adjectives freely: "lo-fi hip hop, boom bap, rainy day, mellow, late night vibes" all contribute.
A few style dimensions worth thinking through before you write your prompt:
- Tempo feeling: slow, mid-tempo, uptempo, driving, laid-back
- Vocal type: male, female, group, no vocals, rap, spoken word
- Instrumentation: acoustic, electric, orchestral, synth-heavy, stripped-back
- Era or genre: 70s soul, 90s grunge, 2010s EDM, classic country
- Mood: melancholic, triumphant, playful, urgent, romantic
You don't need all five in every prompt. Pick the dimensions that matter most for your specific use case.
Custom Lyrics vs. Auto-Generated
When you toggle on Custom Mode in Suno, you get two text fields: one for the style prompt and one for the lyrics. This is where the real control lives.
Auto-generated lyrics are fine for background music or when you just want to hear a genre. But if the song is going into content you're publishing, you almost always want custom lyrics. Auto-lyrics have a tendency toward cliches, and they won't reference your brand, your product, or your specific message.
Writing lyrics for AI generation is slightly different from writing for a human singer. Keep lines short (6 to 10 syllables works well for most genres). Use repetition deliberately, because Suno picks up on repeated phrases and often treats them as hooks or motifs. Avoid lines that are phonetically dense or have complex consonant clusters, because the model sometimes struggles with them.
Structure Tags: Telling Suno How to Build the Song
This is the feature most tutorials gloss over, but it's the one that most changes the output.
Suno reads structural tags inside square brackets and uses them to understand the form of the song. You put these tags at the start of each section in your lyrics field.
Common tags and what they do:
| Tag | Effect |
|---|---|
[Verse] | Standard verse energy, forward-moving |
[Chorus] | Higher energy, fuller sound, repeated hook |
[Pre-Chorus] | Builds tension into the chorus |
[Bridge] | Contrast section, often more minimal |
[Outro] | Signals the song to wind down |
[Instrumental Break] | Generates a music-only section |
[Hook] | Short catchy phrase, often repeated |
A basic song structure might look like this in the lyrics field:
[Verse]
Walking through the city at noon
Everybody moving too fast
[Pre-Chorus]
But I know that feeling won't last
[Chorus]
Just slow down, breathe it in
Start again
[Verse]
...The tags are case-sensitive and need to be on their own line. If you leave them out entirely, Suno will still generate a song structure, but you lose control over where the energy shifts happen.
One thing I noticed in testing: the [Instrumental Break] tag is genuinely useful if you're creating content that needs a musical interlude. A 30-second podcast intro, for example, could use [Intro] followed by an [Instrumental Break] and then a brief [Outro] with a single lyric line. You don't need a full song every time.
Extending a Song
Suno generates songs in chunks, typically around 1 to 2 minutes. If you want a full 3-minute track, you use the Extend feature.
After your first generation, click the three-dot menu on the song and select Extend. You'll see a text field where you can continue the lyrics and style description from where the previous clip ended. You can also select a specific timestamp within the existing clip as the extension point, which is useful if you like the first verse and chorus but want to rebuild from the second verse onward.
Extending is non-destructive: you always get a new variant, and the original clip stays in your library. It's common to generate 4 to 5 extension variants and splice the best sections together in a basic audio editor. Audacity works fine for this; so does GarageBand on Mac.
Regenerating and Exploring Variants
Every time you click Create in Suno, you get two variants by default. They share the same style prompt and lyrics but differ in arrangement, tempo interpretation, and sometimes vocal tone. One might have a heavier bass, the other a cleaner acoustic feel.
Get in the habit of always generating a pair before deciding anything is final. The second variant is often better. When neither is quite right, the trick is to adjust one specific thing in the style prompt (swap one adjective, add one instrument) rather than rewriting everything. Small changes produce meaningfully different outputs.
If you find a variant you almost love but one section is off, use the Cover feature (available on some plans) to remix just that section while keeping the rest.
Stems and Downloads
Suno's paid plans include the option to download individual stems: vocal track separate from the instrumental. This is what you want if you're doing anything beyond casual listening.
Separate stems let you:
- Mix the vocal louder or quieter in your own DAW
- Use just the instrumental under a voiceover
- Apply EQ or effects to the voice without affecting the music bed
- Edit the timing of specific sections without artifacts from combined audio
Download formats are typically MP3. If you're delivering music for professional video production, you'll want to check whether the platform needs WAV; you can use a converter, but note that MP3 is a lossy format, so converting to WAV doesn't recover quality lost in the original encoding.
Commercial Rights
This is where many people get tripped up. Suno's commercial rights work as follows at the time of writing: paid subscribers (Pro and Premier plans) receive commercial usage rights for songs they generate. Free tier users do not.
Commercial rights mean you can use the song in a video you monetize, in a podcast with ads, in a product demo, or in a client deliverable. What you cannot do on any plan is claim copyright over the generated song as a standalone musical composition in most jurisdictions, because current copyright law in the US and EU does not protect works generated entirely by AI without substantial human creative input.
For most practical content creation purposes, commercial rights under Suno's terms are sufficient. For anything going into a major commercial campaign or a product that will be distributed at scale, talk to a lawyer before relying solely on the platform's terms. The legal landscape around AI-generated music is still evolving.
Getting a usable song out of Suno comes down to a specific style prompt, intentional structure tags, and a willingness to generate three or four variants before picking one. The tool produces surprisingly good results when you give it clear direction. Treat the first generation as a rough sketch, not a final answer.