Why does Suno ignore my BPM, and how do I actually lock the tempo?

The complaint is real

You wrote 140 bpm in the prompt. You got a mid-tempo shuffle. You tried fast. Same result. Users say it plainly: "'fast' and '140 bpm' prompts were ignored" and "Suno does not listen to Tempo cues and intentionally slows your song down." One review called it what it feels like: "A music program that doesn't follow BPM is utterly useless."

Here is the honest part first: Suno does not read BPM as a hard parameter. A bare number is a weak signal it is free to round off, average out, or ignore. If you're chasing an exact 128 for a DJ set, no prompt trick guarantees it. What you can do is stack signals so the model lands in the right lane far more often. Think of it as raising the odds, not flipping a switch.

Why the number alone fails

Suno was trained on descriptions of music, not on a metronome. 140 bpm is four characters with no groove attached. The model has no strong link between that string and a felt speed, so it defers to whatever tempo the genre and mood suggested. Give it a genre with a soft average tempo and vague energy words, and it drifts to the middle.

So stop leading with the number. Anchor the tempo through things the model actually understands.

The four anchors

1. A genre reference with a built-in tempo. drum and bass implies ~174. liquid dnb, four-on-the-floor house, boom bap all carry speed in their DNA. Pick a style whose native tempo is your target. Browse the catalog to find one.
2. Feel words, not digits. driving, uptempo, half-time feel, four-on-the-floor, frantic, laid-back. These map to training data. driving four-on-the-floor beats 128 bpm every time.
3. Rhythmic meta-tags, section by section. In your structured lyrics, mark the groove where it lives: [Verse: half-time drums], [Chorus: double-time, driving]. This ties tempo to structure instead of hoping one global number sticks.
4. Name the groove. syncopated 16th hats, driving eighth-note bassline, four-on-the-floor kick tells the model what the pulse should do, which pins the felt tempo harder than any label.

Tempo cheat-sheet: task or genre → BPM range

Use this as a starting lane, not a guarantee. Pick the row that matches your goal, then pair the range with a genre reference and feel words instead of pasting the raw number. These ranges come from recipes we've tested in the builder.

Read a row as a full prompt seed, not a number. For a study playlist you'd write lo-fi hip hop, laid-back, mellow, ~78 feel rather than 78 bpm. For a workout drop: big-room EDM, driving, four-on-the-floor, ~132 feel. The genre carries the speed; the feel words push the energy; the ~N feel hint nudges without pretending to be a hard lock.

Feel words beat a bare number

This is the single highest-leverage swap. A digit is inert; a feel word is loaded with training data. Watch the same intent land two ways:

• Weak: russian rap, 90 bpm — the model shrugs and averages.
• Strong: russian melodic rap, driving, mid-tempo pocket, ~90 feel — the model has three handles to grip.

Keep the feel vocabulary tight and reusable: driving, uptempo, half-time feel, four-on-the-floor, laid-back, frantic. If you must include the number, demote it to a hint (~132 feel) sitting next to a genre, never standing alone as 132 bpm. Explicit-but-supported beats explicit-but-naked, and "mid-tempo" on its own is the worst of all — it tells the model to do exactly what it already does when confused. If you want the RU-vocal version of this same principle, see russian lyrics for Suno.

Per-section rhythm meta-tags

One global tempo has nothing to hold onto across a three-minute arc, so the model flattens it. Attach the groove to sections instead. This is the same move as directorial dynamics — you're directing the arrangement beat by beat.

[Verse 1: half-time drums, laid-back] [Pre-Chorus: build, hats double-time] [Chorus: four-on-the-floor, driving] [Bridge: drop to half-time feel]

Each tag is a fresh anchor, so the model can't average the whole song into one lukewarm shuffle. A half-time verse into a double-time chorus is a contrast the model renders well because you named where the change happens. This is why tempo control and song structure are the same problem — a tempo that isn't attached to a section has nothing to hold onto.

The tempo-lock recipe

Combine all four in one pass:

[Genre: liquid drum and bass, ~174 feel] driving, uptempo, syncopated breakbeat, four-on-the-floor energy in the drop

Then inside the song body:

[Intro: sparse, half-time feel] [Drop: full double-time breakbeat, driving]

Genre sets the lane. Feel words push the energy. Section tags stop the model from flattening everything to one average.

Do this now

• Delete the bare BPM number, or keep it only as a ~174 feel hint next to a genre.
• Pick a genre reference whose native tempo matches your target (use the cheat-sheet above).
• Add two or three feel words: driving, uptempo, half-time feel.
• Tag the groove per section, not globally.
• Build the whole prompt in the builder so the anchors stay consistent across sections.

If tempo is still slipping, the deeper issue is usually how Suno weighs prompts overall — see why Suno ignores your prompt. To make a good tempo repeatable across renders, pair this with reproducible Suno results. More prompt-craft lives in the full guides index and the suno prompts primer; browse styles in the catalog or by artist, and see pricing when you're ready. We can't promise a metronome. We can make the model land on your tempo far more often than a lone number ever will.