How Treblo takes direction
A Treblo generation starts from a prompt, with optional tag and lyric fields. Two dials decide how hard the prompt pushes: prompt strength raises adherence but costs naturalness, and style scale does the same for the tags in the current model. Because the free tier runs unlimited songs and allows using them outside the app, iteration is cheap — which is exactly what these dials reward.
Layer by layer
- Two sentences: the song, then the sound. "A late-night apology song. Warm analog synths, brushed drums, a close female vocal at 92 BPM."
- Leave tags blank at first. Treblo infers them from the prompt, and its guess is often closer than a list you assembled by hand.
- Add tags only to correct drift. If the take came back bright, add "warm, dark, tape-saturated" rather than rewriting the prompt.
- Write lyrics when the words matter. Leave them blank when the hook matters more and let the model find one.
- Move one dial at a time. prompt_strength when the song ignores your description, style_scale when it ignores your tags.
- Extend, do not reroll. Keeping the take and extending it preserves the arrangement you already chose.
Treblo setups you can paste
The prompt alone is enough — add tags or lyrics only when a take drifts.
Indie folk · 92 BPM · G major
Synthwave · 100 BPM · A minor
Solo piano · 70 BPM · instrumental
More ready-made prompts across moods and engines: AI music prompt examples.
What to leave out
- A tag list when the model already inferred well. Every tag you add is another constraint competing with the prompt.
- Both dials at once. Move one, listen, then decide — otherwise you learn nothing about either.
- Artist names. Same trade as everywhere: more risk, less specific instruction.
- Assuming a negative prompt exists. Describe the thing you want present instead.
- Rerolling a take you liked. Extend it.
The rule underneath all of them: describe the music, never the record. Reference an artist or a track and you inherit someone else's rights problem along with a vaguer brief.
Build a Treblo brief from a reference track
The hard part is the honest numbers: tempo, key, brightness, dynamics, and the instruments actually carrying the groove. Music DNA Studio measures them from any MP3, WAV, FLAC or MP4 — locally in your browser, nothing uploaded — then outputs a Treblo-shaped prompt alongside versions for the other fourteen engines. You make the creative calls; the spec underneath is measured, not guessed.
The prompt is doing almost all the work here, so accuracy pays: a measured tempo, a real key and a named instrumentation beat adjectives every time.
Guides for the other engines
Eleven Music · Suno · Udio · Mureka · MiniMax Music · YuE2 · Soundverse · Treblo · Stable Audio · Soundraw · Mubert · Beatoven · AIVA · Loudly · MusicGen
Build The Treblo Brief
Drop in a reference, pick Treblo, copy the prompt and dials. Free, no signup, nothing leaves your device.
ANALYSE YOUR AUDIO — FREE →FAQ
Is Treblo really free and unlimited? ▾
The free tier runs unlimited songs on the current model, with no credits or daily cap, and its FAQ says songs can be used outside the app even on that tier. The API is a separate paid product — check the current terms before planning a release.
Do I need to write tags? ▾
No. Treblo recommends sending the prompt alone and inferring tags, lyrics and strength. Tags are a correction tool: use them when a take drifts, not as a second brief.
What do prompt strength and style scale do? ▾
Both raise how closely the result follows your instructions, and both cost naturalness as they rise. prompt_strength pulls toward the prompt text; style_scale pulls toward the tags. Adjust one per take.
Can I use Treblo songs commercially? ▾
Their documentation suggests yes, including the free tier, but terms change and this is a licence question, not a technical one. Re-read the current terms at release time and keep a record of the date you checked.
How do I keep a direction across several takes? ▾
Hold the prompt steady and change one dial or one tag. That isolates what moved the result — the same discipline as holding a seed in a seed-based tool.