HITL provenance builder.

AI music should not be flattened into one checkbox. Use this page to record human direction, AI assistance, review points, and contribution ratios.

01 What are we trying to understand?

Choose the provenance job first. A finished track, an archive item, and a new-song plan need different questions.

Track or project title Artist name Project stage Working category

Provenance goal

02 What did the human bring before the tools?

Capture intention, taste, lived context, references, and decisions. This is where human authorship becomes visible.

Human intention or creative seed Source material and boundaries Human decisions that shaped the result

03 What roles did AI or tools play?

Pick the roles, then name tools. This helps avoid the vague all-purpose phrase "made with AI".

Idea expansion Lyric drafting or revision Composition or arrangement sketch Vocal, performance, or stem generation Mixing, mastering, or cleanup Artwork, video, or promotional assets Metadata, release notes, or agent workflow Unclear and needs review

Tools used Tool suggestions or AI decisions to review

04 How much did each side contribute?

Use these as honest working estimates. The context above matters more than false precision.

Lyrics human % 85 Composition human % 70 Vocals human % 100 Mixing human % 60 Artwork human % 30 Metadata human % 50

05 How should this be discovered or categorised?

This links provenance to recommendation systems without pretending the platform knows the artist's intent.

Intended listener or scene Discovery lane

Avoid mislabelling as

06 What happens next?

Use this as a planning surface for a new song pipeline or the next agent handoff.

Suggested next pipeline action Disclosure style

Pipeline plan Questions for agent or human review Human review checkpoints Plain-language disclosure