Music Gen is the current name of x-rush.ai's music feature. Some links and older pages still carry its earlier name, Music Studio, this page's address among them. The feature is one product: a generator that turns lyrics and music directions into finished tracks, in two modes, at one fixed price.
Two Modes, One Price
A Music Gen workspace holds exactly one mode. Songs come out with a lead vocal; pure music comes out without one and works as standalone listening or as background for other projects. Both bill the same way: one complete track per generation, 5 credits, with the price on the confirm button before anything runs. There is no second confirmation dialog and no paid preview step. Workspaces live inside a Project, and everything stays private to it until you publish.
Songs
A song needs lyrics and at least one music direction: genre, mood, instruments, structure, and a singing language are the usual fields. Lyrics have three starting points. Ask the copilot for inspiration and take or leave what it drafts. Hand it a theme, a story, or a fragment and let it write a full draft. Or paste your own; pasted lyrics are kept as written, and the copilot touches them only if you request edits.
Before a task is created, the request passes a copyright boundary. Reproducing an existing song, covering a melody, or imitating a recognizable singer's voice stops the task before any credit is reserved. The input is kept so you can rephrase it as era, genre, mood, and voice characteristics and try again. Confirming the generation also asserts you hold the rights to the lyrics.
Pure Music and BGM
Pure music asks for two things: a target duration within the range the interface shows, and at least one music direction. You never need BPM, key, or arrangement parameters; the copilot fills those in as suggestions, and you can overwrite any of them. The generation returns one complete track. Duration is a direction, so the finished file can land off the target; a regenerate is a new 5-credit unit.
The homepage Hero has a BGM shortcut that produces a pure-music result directly, without a workspace. "Continue in Music Gen" attaches that result to a pure-music workspace carrying the final prompt, target duration, and suggested tags. Attaching copies nothing and charges nothing. A mode switch after results exist always starts a new single-mode workspace with the general directions carried over, never the lyrics or previous results.
One workspace runs a single music task at a time; how many workspaces generate in parallel follows your plan: 1 on Free, 2 on Lite, 3 on Pro.
Working with the Copilot
Every Music Gen workspace has a resident Workflow Copilot. It suggests genres, arrangements, and moods, points out missing generation conditions, drafts lyrics in song mode, and completes scene, rhythm, structure, dynamic changes, looping, and ending choices in pure-music mode. Its suggestions are always editable, and it can never start a paid generation on its own or override constraints you set.
Requesting and receiving an inspiration, lyric draft, rewrite, or description polish counts once on the account-wide assistance pool: every 10 successful assists cost 0.5 credits. Typing and editing yourself never advances the counter.
What You Get Back
A result appears only after the full track passes delivery checks: the generation succeeded, the audio decodes with a real duration, and both MP3 and WAV are saved. Failures retry inside the same task, and a track that still cannot be delivered puts the reserved credits back on your balance. Every success is kept; a regenerate adds a new version and never overwrites the old one.
Playback is full-track listening with progress scrubbing. Download delivers MP3 by default with WAV selectable, both of the complete generated audio. There is no editing inside Music Gen: to change anything, edit the input and confirm a new generation. Workspaces autosave from creation, so a refresh or closed tab keeps your lyrics and directions, and a newer save from another tab or device pauses the stale page and offers the latest copy or a save-as-new workspace.
Cover art uses a system cover by default, or any image already in your media library; swapping covers never bills. Publishing to Explore is separate and optional: it fixes a snapshot of the chosen track with title, cover, mode, and duration. Lyrics stay private unless you publish them, the final-prompt and Remix toggles are off by default and independent, and the public detail page offers download and Remix with no copy-link or share-link buttons.
What Music Gen Does Not Do
Every finished track ships as two files, MP3 and WAV, of the complete audio. Around that core, the boundaries are specific:
- No stems or multitrack output. There is no vocal/instrumental split, no independent vocal and backing volume, no mixing session after the fact.
- No vocals-only export. The two modes are the whole output surface: song with lead vocal, pure music without.
- No mixing-desk instrument control. Instruments are one editable direction alongside genre, mood, and structure; they are not channels you balance on a finished track.
- No section-level redo. One generation is one complete track; changing a verse means a new 5-credit unit.
- No reference audio, covers, voice cloning, or MIDI input. Requests in that direction are stopped before credits are reserved.
Delivery is objective, and taste is not part of it. Music Gen does not verify that lyrics were sung word for word, and it does not score genre, mood, or performance. A track that delivers completely settles the generation even if it misses your ear; the remedy is a fresh 5-credit generation with edited input.