Style and lyric direction
Create vocal or instrumental tracks from structured creative input.
Music generation as the beginning of a project
Create a song with Music 3.0 and carry its prompt, lyrics, style, cover, model metadata, private audio, and credit record directly into the MV Studio.
Prompt and lyric direction · private R2 audio · durable Track asset · direct MV handoff
MiniMaxMusic Studio is an independent third-party service and is not the official MiniMax website.
Choose a generated song or upload an authorized MP3/WAV.
Select a 15- or 30-second highlight, output ratio, and visual direction.
Generate and edit a versioned storyboard before any H3 job is submitted.
Attach validated reference images and approve an immutable server quote.
Render each H3 shot independently, review it privately, and rerender only what needs work.
Compose a final H.264/AAC MP4 with the original song as the only master audio.
A generated song is more useful when it becomes a durable product asset. MiniMaxMusic Studio persists the title, prompt, lyrics, style, instrumental setting, model version, output options, duration, cover, credit cost, private audio key, and visibility.
That asset can be played and downloaded through short-lived owner-checked URLs, projected idempotently into the unified Track model, enriched with Track DNA, and selected for a music-video project without re-uploading.
Create vocal or instrumental tracks from structured creative input.
Copy generated audio into owner-scoped storage before marking the task successful.
Compute cost on the server, consume transactionally, and refund generation failures.
Use the same generated song in Create without downloading and uploading it.
The song can continue into analysis, storyboard, H3 shots, review, and final export.
Create a song and its first vertical visual release in one project.
Generate an instrumental and pair it with abstract or cover-led motion.
Use AI content assistance, then preserve the chosen lyrics for final subtitles.
Keep each generated song as a separate private asset before selecting the strongest one.
No. It is an independent third-party product that integrates the model API and provides its own project, storage, billing, and MV workflow.
You select a music range, generate a versioned storyboard, edit every shot, apply validated visual references, approve the storyboard, and request a server-side quote. No H3 job starts merely because a song was uploaded.
Successful private H3 shots are normalized and concatenated by a dedicated FFmpeg compositor. H3 audio is ignored; the selected portion of the original song becomes the only master soundtrack.
The first production formats are a 15-second teaser and a 30-second social clip in 16:9, 9:16, or 1:1. The project architecture is designed to support longer formats later without treating them as one opaque generation.