Short answer: put [Verse], [Chorus], [Bridge] and the other section tags on their own lines. Use commas and periods for breaths, keep lines short, and do not put notes in parentheses because Suno may sing them.
Your lyrics decide the structure of the song. The same block of text behaves very differently with and without section tags - leave them out and the model has to guess where the verse ends and the chorus begins.
Short answer
- Put each tag on its own line, directly above the section it applies to: [Intro], [Verse], [Chorus], [Bridge], [Outro].
- English words inside the brackets are the safest, and capitalization does not matter.
- Number repeated sections - [Verse 1], [Verse 2] - so the model can tell two verses apart.
- Vocals sound wrong? It is usually the writing rather than the model: characters with two readings outside a full sentence, language switching inside one line, and lines longer than about twenty characters.
Structure tags Suno recognizes
Write each tag on its own line, directly above the section it applies to:
| Tag | What it does |
|---|---|
| [Intro] | Opening, usually instrumental only |
| [Verse] | Verse, the narrative section |
| [Pre-Chorus] | Pre-chorus, builds tension toward the chorus |
| [Chorus] | Chorus, the hook of the song |
| [Bridge] | Bridge, creates a turn |
| [Outro] | Outro, the closing section |
English words inside the square brackets are the safest choice, and capitalization does not matter. For multiple verses, write [Verse 1], [Verse 2] to tell them apart.
A complete example
[Intro]
[Verse 1]
Two in the morning at the subway station
The last train takes the final light away
[Chorus]
I fold my worries into paper planes
And let them fly across this city that never sleeps
[Outro]
Once the structure is clear, the model handles the emotional rise and fall between sections on its own, and the chorus is more likely to land as a hook.
Why the vocals come out mispronounced
Most of the time it is not the model - it is how the lyrics are written:
- Polyphonic characters with no context: characters like "行", "还" and "长" can be read more than one way, and on their own the model may pick the wrong reading. Put them inside a full sentence and accuracy improves noticeably.
- Frequent Chinese-English switching: flipping languages back and forth inside one line tends to blur the enunciation. Keep a whole section in Chinese, or a whole section in English.
- Lines that are too long: once a line passes twenty characters, the model often has to compress the rhythm to fit it in, and the delivery sounds rushed. Splitting it into two lines usually fixes it right away.
Punctuation and blank lines
- Commas and periods control where the singer breathes - do not leave a whole section unpunctuated.
- Leave a blank line between sections; it is easier to recognize than running them together.
- Do not put notes in parentheses, or the model may sing the notes too.
A copyright note on lyrics
Lyrics you wrote yourself carry no copyright problem and are safe to use commercially. Using the lyrics of someone else's song directly is a case that may infringe; the platform does not license it, and it is not recommended for any commercial delivery.
¥0.6 per generation, 2 tracks, unlimited downloads
Runs on the Suno V6 engine. Download MP3 or WAV as often as you like. Signing up needs an email address, not a phone number.
