How to Record an AI Music Detector Result Without Turning It into a Verdict

Description and use

This guide explains how to accurately record AI music detector results using the public Creatune AI Music Detector interface and how to distinguish transport failures from musical evidence.

Problem

When a music‑submission workflow includes an AI detector, the application must decide what to store from the detector’s output and how a reviewer should interpret it.

Outcome

The workflow ensures that detector outputs are captured accurately, transport failures are distinguished from musical evidence, and reviewers are guided to seek additional context before making a verdict.

Workflow

• 1. Upload the earliest authorized export of the track to the Creatune AI Music Detector.
• 2. Record the file’s SHA‑256 hash, source description (e.g., mastered release, compressed preview, source export), and the timestamp of the check.
• 3. Copy the displayed score, label, and any duration warning from the public page into a JSON record.
• 4. Store the record with null fields for score/label until the detector returns a result.
• 5. If the detector fails or the request times out, mark the review state as "Request failed".
• 6. If multiple formats (e.g., WAV and MP3) are checked, keep separate records for each hash and export description.
• 7. When a reviewer sees a high score, request additional context such as project files or stems; a low score does not rule out AI generation.
• 8. Maintain a review status that reflects whether the result is recorded, reviewed with context, or still pending.

Evidence

The safe engineering distinction is between a measurement from a particular file and a claim about how a song was made.

Source

After a check, copy the displayed score, displayed label, analysis date, and any duration warning.

Source

The public guide distinguishes likely AI-generated, inconclusive, and likely not AI-generated.

Source