A beat is a rhythm-first instrumental: drums, bass and a repeating musical idea, usually without vocals. This page shows how to get one out of the Lyria 3 workbench, what the prompt needs to contain, and which parts of beat-making the platform does not do.
- Open the AI Music Generator and sign in — every generation is a task tied to your account.
- Switch to Custom Mode to reach the Title, Style of Music, Instrumental and model fields in one form.
- Turn on Instrumental. Leave the lyrics field empty; the form only asks for lyrics when instrumental mode is off.
- Write a rhythm-first style string. Name the genre, the tempo in bpm, and the drum and bass elements you want, for example
trap, 140 bpm, 808 bass, sparse hi-hat rolls, minor keys or boom bap, 90 bpm, dusty kick, swung snare, vinyl texture.
- Check the model and its credit cost — the picker shows the cost of the selected model before you submit.
- Generate and listen for the groove first. Judge tempo and swing before you judge the melody.
- Iterate one variable at a time — change only the bpm, or only the drum description, then regenerate.
- Extend for length. Use Extend with a continue-from point in seconds to add an intro, a switch-up or an outro, then download the finished track.
- Genre plus era. "Trap" and "2010s trap" pull different production styles.
- Tempo in numbers. bpm is the field the model reads most literally.
- Drum and bass specifics. Kick, snare, hats, 808, sub, and how busy they are.
- Sparse beats need sparse words. If you list ten instruments, expect a dense mix.
- A rapper or topliner who wants a bed to write over.
- A video editor who needs an instrumental loop for a cut.
- Practice and teaching: a steady groove to play or count along with.
- No DAW-style editing. You cannot slice, quantise or reprogram individual hits in the platform. If a pattern is wrong, change the prompt and regenerate.
- No default per-instrument stems. The workbench can separate a finished track into two stems or multiple stems as a separate task, but that is not the same as exporting a dry drum stem for your own mixing.
- No MIDI export. The output is audio.
- Not unlimited. Every task consumes credits, and heavier models cost more. There is no fixed number of generations.
- No automatic commercial rights. Commercial rights are not automatically granted; eligibility depends on the selected model provider, your plan, your inputs, applicable law and third-party rights. Check the Terms of Service.
- Asking for an artist's signature sound. Describe the production style instead of naming an artist.
- Leaving tempo out. Without a bpm figure you get whatever the model decides.
- Stacking vocals into a beat request. If you want a beat, keep Instrumental on and the lyrics field empty.
- Regenerating the whole prompt after one bad take. Change one thing per attempt so you learn what the model responds to.
Yes. Enable Instrumental Mode: the task is generated as a vocal-free arrangement, which is the usual baseline for a beat.
Length is decided by the model for the task you submit. If you need more, use Extend in the workbench to continue the track from a chosen point.
Not as a dry drum stem. You can run vocal separation to split vocals from the instrumental, and the workbench offers two-stem or multi-stem separation, but the result is separated audio rather than a DAW-ready drum track.
Each submitted task consumes credits from your account, so yes — attempts are billed. The per-task cost of the selected model is shown before you submit, and Pricing lists the current credit packs and plans.
Open the AI Music Generator →
