Generator Inputs and Style Tags
Every control the AI Music Generator form actually has — the prompt and custom modes, the style tag groups, the sliders and the voice choice — sourced from the form schema, not from marketing copy.
This page lists what the AI Music Generator form takes, read from the form schema that renders it. If a control is not on this page, the generator does not have it. A test pins the list to the schema.
The twelve controls
| Control | What it does |
|---|---|
| model | Which model version runs the generation — five supported versions for new songs. The version sets the input limits, which of the controls below appear and how many songs a run returns. |
| instrumental | A switch. On, the track comes back with no vocals. |
| prompt | In prompt mode, a description of the track, with a Get Inspired fill and the four style-tag groups as quick-add chips. In custom mode, the lyrics themselves. The limits are the version's (Formats and Limits). |
| title | A name for the track in custom mode. |
| style | Genre, mood, tempo and voice tags, typed or picked from the tag groups below, with a Get Inspired fill of a complete tag line. |
| negativeTags | Styles to steer away from. Custom mode only; not on Music Solo. |
| vocalGender | Random, male or female vocal. Custom mode only, hidden for an instrumental, and only on Music V1, V2 and V2.5. |
| styleWeight | How strongly the style tags steer the result, 0 to 1. Custom mode only, on Music V1, V2 and V2.5. |
| weirdnessConstraint | How far the result may drift from the expected form, 0 to 1. Custom mode only, on Music V1, V2 and V2.5. |
| duration | How long the song runs, any whole second from 3 seconds to 10 minutes. Music Studio only, in both modes; its price follows the length. |
| visibility | Public or Private, in both modes. Private needs a paid plan, and a paid account starts on it; without one, songs are public. Public songs may appear in public lists, though no page lists them yet. A share link works either way. |
| audioWeight | How closely the result follows source audio, 0 to 1. Shown only where a song or upload is the source (extend and cover), never on a new song. |
Prompt mode uses prompt alone: the model writes the lyrics, title and arrangement from your description, and a chip clicked under the description appends that tag to the description rather than filling a separate field. Custom mode uses prompt as lyrics plus title and style, the exclude-styles box, the vocal-gender control and the style and weirdness sliders. There is no upload control on this form: a source song is chosen from your library, not uploaded here.
Style tags
The style field offers 69 tags in four groups, and a generator page without a studio profile shows those four groups. A genre page with a studio profile (the catalogue in music-studio-profiles.ts — hip hop, techno and ambient shipped first) replaces them with its own chip rows, adds a preset row that fills the whole tag line in one click, and adds tempo, key and texture controls. In prompt mode the same groups sit under the description as quick-add chips.
| Group | Tags |
|---|---|
| genre | 29 |
| moods | 18 |
| tempos | 11 |
| voices | 11 |
Naming a tempo in beats per minute in the prompt is more reliable than naming a style that implies one.
What the controls cannot do
The sliders steer; they do not guarantee. A high styleWeight narrows the result towards the tags, and a reference track with a high audioWeight narrows it towards that recording — neither produces a finished master. Mastering is a separate tool, and commercial-use rights come from a licence, not from any control (see Commercial license).
Next: Formats and Limits or Credits and Plans.