How to export your lyrics from Suno
You can regenerate audio from a prompt. You cannot regenerate a prompt from silence. Since September the words are the part of your catalogue you can actually take with you.
Last updated
Why the words are the part worth saving
Suno's cap counts audio files. It does not count anything else. Your lyrics, the style prompt that produced the sound, which model made it, when you made it and how often it has been played are all sitting in your own browser every time you open your library, and none of it is limited.
For a library of a few hundred songs, that is the difference between keeping everything and keeping twenty things. The audio for most of your catalogue is staying where it is. The writing does not have to.
Doing it by hand
- Open the song. In your library, click the song title to open it. The lyrics panel is on the song page, not in the row.
- Copy the lyric sheet. Select the whole lyric block including the section markers in square brackets. Those markers are instructions to the model, and losing them means losing half the prompt.
- Copy the style separately. The style prompt lives in its own field and is not part of the lyrics. Copy it too. It is the part nobody can remember six months later.
- Paste somewhere that keeps line breaks. A plain text file or a Markdown document. Paste into a rich text editor and the section markers usually survive; paste into a chat box and they usually do not.
That is about two minutes a song, which is fine for five songs and completely unworkable for five hundred.
Three things that go wrong
Section markers get flattened
The [Verse], [Chorus] and [Bridge] markers are not decoration,
they are part of what you told the model. Paste a lyric sheet into something that strips formatting
and you can lose the structure while keeping the words, which makes the prompt much less useful to
reuse later.
Line breaks arrive as literal characters
Lyrics travelling through an API carry their line breaks as an escape sequence rather than as
real newlines. Copied out of the wrong place they arrive as one continuous block with
\n scattered through it. Any export worth having converts those back into real lines
before writing the file.
The style prompt gets left behind
It is a separate field, so a copy of the lyrics is only half the prompt. Six months later the lyrics will still make sense on their own and the style will be the thing you cannot reconstruct.
Doing it at scale
Suno Manager reads your library as you scroll it and keeps a copy in your browser: title, style, model, creation date, play and like counts, and the full lyric sheet with its markers and line breaks intact. Nothing is fetched behind your back and nothing is sent anywhere.
Once it is archived, three exports come out of it:
- Markdown for reading and printing. One section per song with the style, the model and the lyrics laid out as real lines
- CSV for a spreadsheet, if you want to sort by style, model or play count
- JSON for a complete backup. It is the only one of the three that can be loaded back in, so it is the one to keep somewhere safe
You can also search the archive by a line you half remember, which is the fastest way to find a song when you have forgotten what you called it. The extension is free, and there is more on what a good archive contains in the backup guide.
Keep reading
- Cleaning up your library Eleven takes of one idea is normal. How to find the one worth keeping before you spend a download on the wrong one.
- Which model made this song Suno says v5. The data behind the song says something else entirely. What the two names mean and where to read them.
- Backing up your library Archiving everything that is not audio, and the three export formats worth keeping a copy in.