How to choose the vocalist voice for an AI-generated track

Pick Female or Male under Voice in the studio, then click Voice options to set up to three Character chips, one Delivery, How much vocal, Accent and Tone. Club Ready Music turns those choices into one sentence that starts "Vocals:" in the prompt sent to the music model, such as "Vocals: smooth, soulful female lead vocal, sung, performing the provided lyrics, full lead vocal through verses and chorus, British accent, clear and upfront in the mix."

Steps

1. Choose Female or Male

The two buttons under Voice set the gender sent to the model. On a phone this is on step 3 of 5 under the lyrics. If you later pick a cloned voice, it overrides this choice.

2. Open Voice options

Click the Voice options button next to the gender buttons. A badge on the button shows how many choices you have made. The dialog has six groups and a cloned voice section.

3. Pick up to three Character chips

Choose from Smooth, Raspy, Breathy, Powerful / belting, Airy / whisper, Deep, High / falsetto, Soulful, Gritty and Sweet / pop. Pick traits that go together: Airy and Breathy for deep house, Powerful and Soulful for a big vocal house record, Deep and Gritty for techno spoken lines. Three contradictory chips, such as Airy with Powerful, give the model mixed instructions.

4. Set Delivery and How much vocal

Delivery is one of Sung (default), Rap, Spoken word, Chant / crowd shout, Call & response or Harmonies / choir. How much vocal is Full lead, Hook only or Sparse chops. Sung plus Full lead is a song; Chant plus Hook only is a DJ tool with a shout; Sparse chops is a texture.

5. Choose Accent and Tone

Accent offers Neutral, American, British, Caribbean, Spanish-tinged and French-tinged. Tone is Young or Mature, and it is written in front of the gender, so "mature male lead vocal". Leave Accent on Neutral unless the genre asks for one: Caribbean suits dancehall and UK garage, Spanish-tinged suits Latin house.

6. Tick effects and close

The Effects checkboxes add Ad-libs around the drops and an Auto-tune effect. Click Done. The summary line at the bottom of the dialog repeats your choices in plain words so you can check them.

7. Generate and compare takes

Press Generate. The same voice direction is used for both takes, but the model performs each take separately, so compare them before you download.

Tips

• Start from the genre. Afro house: Smooth, Soulful, Mature, Hook only. Tech house: Breathy, Sparse chops. Big room: Powerful / belting, Full lead. Drum & bass: Airy, Hook only, British.
• Fewer chips are clearer. One or two character traits give a more consistent voice across takes than three.
• Harmonies / choir is a delivery, not a layer. It asks for stacked harmonies throughout. For a lead with occasional harmonies, keep Sung and write "(harmony)" lines in the lyrics.
• The vocal is always mixed upfront. Every direction ends with "clear and upfront in the mix". For a buried, effected vocal, choose Sparse chops and describe the effect in the vibe text.
• Settings are remembered. The studio keeps your voice choices in the browser, so the next track starts from the same singer.

Common problems

The voice sounds different between Take 1 and Take 2. Each take is a separate performance. Choose the take you prefer and render again with the same settings if you need a matching second track.

I picked Male but got a female voice. Check the Cloned voice select in Voice options; a cloned voice overrides Female / Male. Set it to None to use the gender buttons.

The accent is barely noticeable. Accents are a light hint to the model. Writing lyrics with the spelling and phrasing of that accent has a stronger effect than the chip.

Rap delivery still sings the chorus. Rap delivery asks for rapped verses with a sung hook by design. For no singing at all, use Spoken word and keep the chorus lines short. See rapped or spoken vocals.