Suno’s new Speech beta generates a voice and a music bed as one track. That means one prompt has to cover three jobs: the words, the voice and the music. This guide is our own method, so test it on your tool.
The three parts
- The words. Paste your poem, script or idea. Keep sentences short and mark the pauses you want with line breaks.
- The voice. Describe age range, tone, pace and accent in plain words, for example “calm, low, unhurried, slight smile”. Do not ask for a real person’s voice.
- The music. Name genre, instruments, tempo and how loud it sits under the voice, for example “soft piano and warm pads, slow, always quiet under the voice”.
Template
Text: [your words, one idea per line]
Voice: [tone, pace, accent, mood]
Music: [genre, instruments, tempo, volume under the voice]
Ending: [fade out / one last note / silence]Fixes for common problems
- Voice drifts or changes accent: simplify the voice description and name one accent only. Suno itself says the beta can drift on accents.
- Pauses feel too long or too short: add or remove line breaks, then regenerate.
- Music overpowers the words: say “sparse, quiet, supporting the voice” and drop busy instruments.
Build a base with our guide to AI music prompts and see the AI voice prompts page. If you plan to publish, check the tool’s current terms on rights, and only use voices you have permission to use.