Mureka is the fastest-growing song generator after Suno. Its traffic grew by about half in a single month this autumn, and its V9.5 model (August 2026) is built around the same two inputs Suno uses: a description of the music and your lyrics. If you can write a good Suno prompt, you're most of the way to a good Mureka prompt. The rest is knowing where the two differ.
One caveat before the details. Mureka's own developer documentation didn't load when we checked, so the limits below come from the API hosts that resell Mureka (fal, eachlabs, WaveSpeed) and from Mureka's site. Where they disagree, we say so and give the safe number.
The two fields that matter
The style prompt describes the music, not the words. Mureka's own FAQ asks for genre, mood, tempo, instruments and what the song is for, and says more specific descriptions work better. Guides from API hosts add the vocal and the song's shape: where it stays sparse and where it opens up.
The lyrics carry the structure. Square-bracket section tags such as [Verse], [Pre-Chorus], [Chorus] and [Bridge] are honoured, and the API hands back where each section landed. Mureka publishes no full tag list, so stick to the common section names.
A style prompt in that order reads like this:
Indie pop, 112 BPM, bright and hopeful. Early-2010s bedroom-pop warmth.
Warm female lead, close-miked and conversational. Brushed live kit, round
electric bass, chiming clean guitars and a glockenspiel line. Verse 1 close,
conversational; Chorus full band, belted; Bridge half-time, whispered.
Vocal up front, guitars wide. Ends on a held chord ringing out.
Genre and tempo first, then era, the voice, the band, how each section is delivered, the mix and the ending. It's the same order that works in Suno's Styles field, for the same reason: the start of the prompt carries the most weight.
The limits
- Style prompt: about 1,000 characters when you also give lyrics. Hosts list 1,024; one developer doc says 1,000. Aim under 1,000. A prompt-only song (no lyrics) gets more room.
- Lyrics: 3,000 characters is the safe figure. Some hosts allow 5,000, others stop at 3,000, so a sheet longer than 3,000 may be cut short.
- Length: 2 to 5 minutes, set as a duration, not written into the prompt.
- Vocal gender: a setting (female or male), not a prompt word. For a duet, leave it unset and describe both voices in the prompt.
- Reference audio: a 30-second reference clip, a voice clip or a melody (humming or MIDI) can steer the song. The melody input overrides the rest. Reference uploads are a Pro feature on the site.
Lyrics Mureka can read
Write the sections out in full, with plain tags:
[Verse 1]
You left the porch light on for me
I counted every car that passed
[Chorus]
Leave the light on, leave the light on
I'm halfway home (halfway home)
[Bridge]
If the road goes dark
I'll follow the glow
[Chorus]
Leave the light on, leave the light on
I'm finally home
[Outro]
A few habits help, and most of them are the same ones that help in Suno:
- Repeat the chorus word for word each time it comes back. Guides from Mureka's API hosts make the same point: the model holds a chorus better when its text matches.
- Keep matching sections the same shape. Verse 2 should have about as many lines, and lines about as long, as verse 1, so the melody can repeat.
- Keep each section on one idea. A verse that tells one moment sings better than one that lists four.
- Use parentheses only for what you want sung as a backing part, like the echo above. Directions in parentheses, like "(softly)", may be sung too.
Bringing a Suno prompt to Mureka
Mureka takes the same two things Suno does, so most of a Suno prompt moves across. Three things change.
Take the cue out of each header. Suno reads [Chorus | full band, belted]; Mureka's guides only show plain [Chorus], and nothing says what Mureka does with the words after the pipe. The direction isn't lost if your Styles field already says how each section is delivered (the "Verse 1 close, conversational; Chorus full band, belted" sentence above). Inline cues like [beat stops] and voice tags like [Male vocal] should come out too, for the same reason: they may be sung.
Leave Exclude behind. Mureka has no field for what to keep out. Don't turn the list into "no trap hi-hats, no autotune" sentences: none of Mureka's guidance covers negatives, and in Suno, at least, naming a sound in the prompt tends to add it.
Move the settings to settings. Vocal gender and length become Mureka's own controls. A Suno [End | held chord, ring out] line becomes [Outro], with the ending described in the style prompt.
The free Suno prompt converter does all of this for you: paste your Suno Styles, Exclude and lyrics, pick Mureka, and it gives you the style prompt, the tagged lyrics and the settings, with a list of what it changed. The example on this page is its output.
What we haven't tested yet
We haven't run a blind listening test of converted prompts on Mureka against the same prompts on Suno. The format above follows what Mureka and its hosts document; how closely V9.5 follows a section's delivery is something to judge by ear. Generate two takes (the API makes up to three per run) and keep the one that follows the plan.
If you're still choosing between the two tools, our Suno vs Mureka comparison covers price, rights and what each does best. And whichever you use, the free lyric checker catches the lyric habits that make any model sound generated: clichés, lines too long to sing and choruses that drift.