Skip to main content

The Contribution Process

This page describes the mechanics of contributing: what to settle before your first upload, what the app accepts, and what happens to your file once it enters the ingestion pipeline. The same mechanics apply to every contribution, from a single demo to a batch of stems and MIDI files.

If you have not joined yet, see How to Join. For how uploads translate into points and royalties, see How Your Music Is Evaluated and CRPS — Your Stake. For the longer-form take on what the upload moment means — and why CORPUS treats each one as a deliberate act rather than a frictionless flow — see Every Upload Is a Decision in the journal.

Before you upload your first contribution​

Two things to settle before you upload anything you want scored.

Splits and consent. Every work in CORPUS enters with a collaboration agreement that names all rights holders, their percentage split, and an explicit consent to the use of the work as AI training material. Each named party is invited by email to confirm their share. The agreement is captured as an immutable snapshot at the moment of upload. If you change a template later, past uploads stay on the version they were uploaded under.

Vocals. If a track contains singing, the singer must be named in the agreement and must have explicitly consented. Personality rights sit above copyright in this area, and we make no exceptions. Tracks with vocals from unnamed or unconfirmed singers cannot enter the library.

For instrumental work and for tracks where you hold all rights yourself, the agreement reduces to a single confirmation step.

What to upload​

The contribution app accepts:

  • Original recordings. Uncompressed formats (WAV, FLAC, AIFF) score higher than lossy formats. MP3 is accepted for cases where the original is no longer available.
  • Stems. Separated tracks earn additional points, because they are more useful for training. Stems are checked against the master through waveform analysis to confirm they belong to the same work.
  • MIDI and metadata. If you have MIDI files, session metadata, or your own annotations to share, the app accepts these alongside the audio.

How these inputs translate into your score is described in How Your Music Is Evaluated.

What happens after upload​

Every file passes through an ingestion pipeline before it enters the library. The pipeline runs automatically, and you see each step's result in your dashboard.

  • Music detection. Non-music uploads (field recordings, broken files, spoken word) are filtered out. If your work is noise music, experimental, or otherwise edge-case for a music classifier, you can appeal and we will disable that specific filter for your account.
  • Duplicate detection. Audio fingerprinting catches re-uploads of the same recording. If the duplicate is yours (an upload that happened twice), it resolves quickly. If someone else uploaded the same recording first, the conflict goes to human review with a full audit trail.
  • Cover detection. A separate system checks whether your recording is a cover of an existing song. The current database covers roughly thirty thousand of the most-covered tracks. Uncleared covers are held until the rights situation around the underlying composition is resolved.
  • AI detection. Synthetic-audio screening removes machine-generated material. The detector currently catches output from Suno through its second-most-recent model version. It is a checkpoint, not a guarantee; if something slips through and is flagged later, the upload is removed and the contributor is contacted.
  • Stem verification. Uploaded stems are checked against the master by waveform analysis.
  • Vocal and personality rights check. Tracks with vocals are matched against the named singer in the collaboration agreement. Missing consent blocks the entry.

If any check flags your upload, you receive a message in the app explaining why and what to do. Most flags resolve through clarification or a follow-up upload.