Audio-led arrangement context
The source recording gives the model phrasing and timing cues that a text prompt alone cannot carry, especially when the source has a clear pulse and hook.
Build supporting music around vocal or source audio by uploading a recording, describing the arrangement, and generating an instrumental version for review.
Review the source audio, track direction, available controls, and account requirements before you begin.
The workflow keeps the source recording, arrangement brief, and generated backing track together so you can judge whether the new music supports the original performance.
The source recording gives the model phrasing and timing cues that a text prompt alone cannot carry, especially when the source has a clear pulse and hook.
Describe instruments, genre, energy, and production traits while excluding unwanted directions, such as 'warm piano and brushed drums, no EDM drop.'
Review returned audio beside the source and keep task metadata visible during iteration.
This focused workflow is more useful than a generic generator when the task, source requirements, and review step are explicit.
Let an existing vocal or musical idea lead the arrangement rather than forcing it into a generic template.
Compare instrumentation and energy choices before committing to a full production session.
Generated backing music remains a new asset, leaving the original recording untouched.
Upload a clear source recording, describe the role the backing track should play, then audition the result beside the original before using it as a reference.
Choose a clear vocal or musical recording that communicates the timing and phrasing to support.
Set a title, style tags, exclusions, model, and advanced tuning. State the desired instruments, energy arc, and what should stay out of the arrangement.
Check how well the instrumental supports the source, then refine tags for another pass.
These tasks use the output as an arrangement reference: audition it against the source, identify what works, then take that decision into production or another generation pass.
Build a musical bed around an existing vocal sketch or topline before scheduling players, recording sessions, or a full production pass.
Compare sparse, cinematic, electronic, or band-led treatments before production.
Turn raw recorded ideas into clearer references for collaborators and editors.
A supporting track should respond to the timing, phrasing, and emotional shape already present in the uploaded recording.
Choose a recording where the intended melody and pulse are easy to hear. Remove long silence, unrelated conversation, room noise, or competing playback when possible. The goal is not studio polish; it is a source that communicates entrances, phrase length, dynamics, and the part the backing arrangement needs to support without forcing the generator to guess.
Name the rhythm section, harmonic texture, lead instruments, energy curve, and production era when useful. Explain what should stay absent, especially sounds that would compete with the source. “Sparse piano, soft bass, brushed drums, space after each vocal phrase, no synth lead” gives more actionable direction than listing several genres without assigning musical roles.
First check whether the backing track follows the source timing, leaves room for important phrases, and develops energy in the right places. Then judge tone, instrumentation, and mix character. A stylish result that ignores the performance is not a useful accompaniment. Record the exact problem before changing tags so each new version tests one arrangement decision.
Direct answers about source requirements, output, limits, and downstream use.
Move the current result into a relevant generation, editing, separation, or conversion task.
Turn an instrumental into a vocal song by uploading the backing track, describing lyrics and delivery, choosing a style, and generating a new version.
Separate vocals from a stereo song, preview both outputs, and download vocal and instrumental WAV files in a short local workflow.
Turn a clear melody, vocal, bass, or lead recording into editable MIDI notes and download a standard .mid file.
Upload the performance that should lead and describe the arrangement it needs.
Generate supporting tracks