AI generated music is audio produced by a model rather than performed and recorded by musicians. You describe what you want, the model produces a track, and you decide whether to keep it. This page explains the mechanics in plain terms, then shows the shortest path to making one.
- You describe the target. A text prompt sets genre, mood, tempo and instrumentation. Optionally you add your own lyrics, or upload an image, video or audio file so the model can analyse the mood.
- You pick a model. Different music models respond to prompts differently and carry different credit costs. The picker labels the provider and the cost of each option.
- The platform submits a task. Generation runs server-side, so the workbench polls for the result; the progress panel warns that a task may take a couple of minutes.
- The model returns audio. You play it in the workbench, and a finished track can be downloaded from the result panel.
- You iterate or edit. Regenerate with one changed variable, extend the track, add instrumental or vocal layers, or separate vocals into stems.
- The output is a new recording, not a remix. It is not assembled from clips of named songs.
- Text is a direction, not a script. Length, mix balance and exact instrumentation are decided by the model for that task.
- Watermarking is provider territory. Some providers attach provenance and watermarking signals to generated audio. Our own explainer on how Google watermarks AI-generated music covers the SynthID approach; check the documentation of the provider whose model you used for what applies to your file.
- Rights do not come with the file. Commercial rights are not automatically granted by this platform. Eligibility depends on the selected model provider, your plan, your inputs, applicable law and third-party rights.
- Open the AI Music Generator and sign in — an account is required before a task can be submitted.
- Choose text, lyrics or media input depending on what you already have.
- Fill in genre, tempo and instrumentation rather than a single adjective.
- Select the model and check the credit cost shown against your balance.
- Generate, listen, then keep or change one thing.
- Export the track and, if you are publishing, generate cover art in the Song Cover Generator.
- Provider terms for the model you used, plus the platform Terms of Service.
- Platform disclosure rules where you upload: some services require AI content labels.
- Your inputs. Do not submit lyrics, voices or samples you do not have the right to use.
- Provenance. If you need a clean chain of custody, keep the task ID shown in the workbench result.
- Reading "AI" as "free and unlimited". Each task consumes credits shown in your account; there is no unlimited tier and no fixed task count, because per-task cost varies by model and options.
- Comparing one model's output to another's prompt. Prompt wording is model-specific; a prompt that shines on one model can underperform on another.
- Skipping the listen-through. Generated audio can contain artefacts; use negative tags and regenerate rather than shipping the first result.
- Assuming ownership. Check rights before monetizing.
The models produce new audio in response to your prompt; the platform does not stitch existing recordings together. If you upload reference material, you are responsible for having the right to use it.
Provenance tools and watermarking research exist — see our SynthID explainer — but detection is not a guarantee in either direction. For published work, disclosure plus accurate metadata is the practical approach.
Yes. You must create an account or sign in before generating, and every submitted task uses credits that the workbench displays before you submit.
Start with the AI Song Writer workflow: draft words, fix a structure, then generate. If you already have lyrics, go straight to Lyrics to Song AI.
Open the AI Music Generator →
