Generation is the first half of the job; the second half is changing what you already have. The Lyria 3 workbench ships a set of editing tools that run as tasks on a finished track. This page maps each one to the problem it solves.
- Extend — continues an existing generated track. You pick the source track, set a continue-from point in seconds, and describe the next section. Use it for a longer intro, a switch-up or a proper outro.
- Add Tracks — layers new material onto a track: add an instrumental part, or add a vocal layer with its own style prompt and vocal direction.
- Vocal Remover — separates a generated track into vocal and instrumental parts, with two-stem or multi-stem output.
- Cover Art — generates artwork for the track inside the same project, so the release assets are not scattered across tools.
- Voice Cloning — builds a voice ID from an uploaded sample so vocal tasks can reuse it. The tool shows its own requirements and cost before you submit; only use a voice you have the right to use.
- Negative tags — steers the task away from artefacts you do not want, such as noise or harsh distortion.
- Generate a base track in the AI Music Generator and keep the task in your history — the editing tools operate on existing tracks.
- Decide what is wrong. Length, missing part, muddy vocal, or no artwork. Each problem maps to a different tool.
- Extend if it is too short. Enter the continue-from second and describe the next section; be specific about instrumentation rather than repeating the whole brief.
- Layer if a part is missing. Add Instrumental for a new part, Add Vocals for a vocal layer. Both need their own prompt and tags.
- Separate if you need stems. Run the Vocal Remover and choose two stems or multiple stems depending on how much control you want.
- Generate cover art while the audio is final, so the artwork matches the finished mood.
- Export the results. Download the revised audio and artwork from the result panels.
- Too short, or the ending is abrupt: Extend.
- The arrangement is thin in one register: Add Tracks.
- You need an instrumental for a video or a practice track: Vocal Remover.
- You are publishing and need a thumbnail: Cover Art.
- You have a consistent vocal identity across tracks: Voice Cloning.
- It is a task-based editor, not a timeline DAW. There is no fader automation, crossfade, tempo map or per-note editing. Download and finish in a DAW if you need that.
- Every tool run is a billed task. Costs vary by tool and model and are shown before you submit; extending a track is not free just because the source exists.
- Separation is not transcription. Stems are audio, not MIDI, and results depend on the source mix.
- No automatic commercial rights. Rights are not automatically granted and depend on the selected model provider, your plan, your inputs, applicable law and third-party rights — review the Terms of Service.
- Voice cloning has an obvious boundary. Only use a voice sample you have the right to use; cloning another person's voice without permission is not a supported use.
- Extending without describing the next section. A vague continue prompt produces a vague section.
- Running separation on a rough draft. Fix the arrangement first, then split it — separation is another billed task.
- Stacking edits without keeping a good take. Download the version you like before you run the next tool over it.
- Assuming an edited track is a new song for free. It is a new task with its own credit cost.
The editing tools operate on tracks in your workbench. If you have outside audio, use the upload path where the form offers it, and check the tool's own limits before you rely on it.
Only if you describe it. The continue prompt and style fields drive the new section, so repeat the key instrumentation and mood if you want continuity.
Yes — that is what the Vocal Remover is for. Choose two stems for a vocal/instrumental split, or multiple stems when you want more granular parts.
Costs vary by tool and model and are displayed before you submit. Current packs and plans are listed on Pricing, and refund eligibility is described in the Refund Policy.
Open the AI Music Generator →
