Add AI vocals to your beat
Upload a beat and Lyro adds a vocal to it in one of two ways. With lyrics, it generates a sung performance at your beat’s tempo and key, separates the voice and returns it as an isolated vocal stem, so your beat is never altered. Without lyrics, ACE-Step 1.5 listens to the beat and improvises a topline over it. Either way, the beat and the new vocal open together in the mixer.
How it goesThree steps
- Upload your beatLyro detects its tempo and key in your browser in a few seconds. Correct them if you know the real values.
- Give it lyrics, or let it improviseWith lyrics, Lyro sings them at the tempo and key of your beat. With none, it improvises a vocal melody that follows the track.
- Mix the vocal with your beatBoth parts open in the mixer, lined up on bar one. Balance them, add reverb and delay, then master.
The details
The two paths, side by side
| With lyrics | Without lyrics (improvised) | |
|---|---|---|
| What happens | A sung performance is generated at your beat’s tempo and key, then the voice is separated out | The model listens to your beat and improvises a vocal line over it |
| Model | Lyria 3 Pro by default, then stem separation | ACE-Step 1.5 |
| What you get | An isolated vocal stem singing your words | An isolated vocal that follows the music, not written lyrics |
| Cost | 4 credits for the generation plus the separation (one credit plus two per started minute): 11 in all for a three-minute beat | One credit per started minute, minimum 2 |
| Best for | A finished song with words that matter | Topline ideas, hooks and ad-libs |
How the lyrics path works under the hood
The model that sings lyrics best generates voice and music together, so Lyro works with that. It generates a full song at the detected tempo and key of your beat, then runs stem separation on that song and keeps only the vocal. On this path your beat itself is not sent to the music model. Only its tempo and key are used.
One consequence is worth knowing. The vocal is in the same tempo and key as your beat, but it was sung over generated music, not over your exact chord changes and arrangement. Most of the time it sits well. Sometimes a phrase needs moving, trimming or a small pitch shift in the mixer.
How the improvised path works
ACE-Step 1.5 accepts audio as input. It hears your beat and generates a new vocal track that follows its rhythm and harmony, returned on its own so you can mix it. You can describe the voice you want, such as a soulful lead with a melodic hook. It does not follow written lyrics, so use it for melody ideas, hooks and ad-libs. If you like the melody, you can write words to it and run the lyrics path afterwards. A topline is the vocal melody written over an instrumental.
Where the limits are
- AI vocals can mispronounce words, and a new attempt costs credits.
- The isolated stem comes from separation, so it can carry faint artefacts from the music it was sung over.
- Tempo and key detection are estimates. A half-time or double-time reading, or a relative key, will put the vocal out of step with your beat, so check both before you generate.
- Beats with a drifting tempo, such as live recordings, are harder to match than programmed ones.
- Long uploads take longer to analyse and process.
- You need the rights to the beat you upload. If you leased it, check what your lease allows.
Formats, credits and rights
Bringing your own beat is part of the Creator plan, which also includes the mixer, mastering, stems and WAV export. If any step fails or times out, every credit for the job comes back automatically, including the credits prepaid for the separation step.
As between you and us, you own whatever rights exist in the result, and your beat stays yours. Every paid plan includes a commercial licence for the generated vocal. Vocals from Lyria 3 Pro carry Google’s SynthID watermark, and downloads carry an ID3 tag declaring them AI-generated. See music rights and the step-by-step guide How to add AI vocals to your beat.
Questions
Will Lyro change my beat?
No. Your beat is analysed for tempo and key and is used as a reference. The vocal comes back as a separate, isolated stem, and the two are combined only in the mixer, where you control the balance.
What does it cost to add a vocal to my beat?
With lyrics, 4 credits for the Lyria 3 Pro generation plus stem separation (one credit plus two per started minute): 7 credits for a beat of up to a minute, 11 for a three-minute one. Improvised, one credit per started minute with a minimum of 2. Both need the Creator plan, from $19 a month for 300 credits. See the pricing page.
Which file types can I upload?
WAV, MP3 and M4A all work. Upload the best quality you have, ideally the WAV export from your music software, with no limiter pushing it into distortion. You can check its tempo and key first with the free BPM and key finder.
Can the vocal follow a melody I already have in mind?
Not note by note. You steer the vocal with the lyrics, the voice type, the language and the style prompt, and the model writes the melody. If you need an exact melody, record it yourself in the browser and let Lyro build on your voice instead.
Do I need the rights to the beat?
Yes. Only upload audio you made or have permission to use. If you leased the beat, your lease decides what you may release. Lyro’s licence covers what Lyro generates, not the audio you bring. See the acceptable use policy.
Related: AI vocal generator, Mixing and mastering, AI lyrics generator