Audio stays on the computer
The detector is bundled with the application. There is no cloud analysis service, API key, account, subscription, or audio upload.
Three clear assessment outcomes
Results use Likely authentic, Inconclusive, or Likely synthetic rather than forcing a misleading real-or-fake answer.
See the supporting result
Review synthetic probability, confidence, detector identity, model version, summary, and suspicious-segment timeline.
Current detector scope
Generated-music screening is implemented today.
The bundled music model currently targets indicators associated with Suno up to version 5 and Udio up to version 1.5. Synthetic-speech inference is a future milestone and is not represented as a current feature.
Add files or folders
Select individual audio files, scan folders recursively, or use drag and drop to build the analysis queue.
Run local analysis
A bundled local process analyses supported audio without sending content or fingerprints across a network.
Review probability and confidence
Use the assessment, confidence, model information, and timeline as screening evidence for a human decision.
Export the record
Create a versioned JSON forensic report with the result and an explicit evidentiary disclaimer.
Important limitation
Detection can inform review; it cannot establish provenance.
- Compression, transcoding, mastering, denoising, broadcast processing, and re-recording can affect a result.
- An unseen generator or mixed authentic and synthetic material may produce an inconclusive or incorrect assessment.
- Keep the original file and supporting context when provenance matters.
- Do not use the result alone to allege authorship, misconduct, or a rights violation.
Use it as a screening tool