Login

MR Maker quality review

How we evaluate vocal-removal results

MR Maker publishes the checks used for uploaded audio and generated instrumentals. This is a reproducible review process with clear limitations, not a promise of perfect separation.

Last reviewed:

Scope of this checklist

This is a public checklist for future quality research and manual evaluation of authorized samples. It is not an automated promise that every customer job receives full-file decoding, beginning/middle/end listening, or an evidence record.

01

Validate the source file

We check the supported format, file size, duration, and whether the audio can be decoded. Audio that the user is not authorized to use is not eligible.

02

Compare matching passages

We compare the same passages at a similar listening level for lead-vocal residue, backing-vocal bleed, drum or bass damage, and changes in stereo width.

03

Check technical integrity

We verify that the result plays through, preserves a plausible duration, and contains no unexpected silence, clipping, or broken encoding.

04

Pass or retry

A job is complete only after file generation and basic integrity checks pass. Failed processing does not consume credits; a different quality option or a better source file may be recommended.

Known limitations

  • When voices and instruments share frequency or spatial information, vocal residue or instrument damage can remain.
  • Live recordings, strong reverb, lossy compression, and choirs are harder to separate.
  • MR Maker does not grant copyright or commercial-use rights. Users must verify rights for the source and output.
  • This page does not claim a public benchmark score. Results vary with each source recording.

Public test protocol

The review criteria are also published as machine-readable JSON. It contains the test stages, evidence fields, outcomes, limitations, and sources—not an unsupported quality score.

Open the JSON protocol

Protocol version 1.0

Source record
Format, duration, and full-file decode result
File integrity
Full-file decode, output duration, and silence or clipping review
Matched listening
Beginning, middle, and ending passages with review notes
Recorded outcome
Pass, manual review, or retry with a reason and review timestamp

Public benchmark status

No public aggregate score

MR Maker does not currently claim an aggregate benchmark score. A score will be published only when a licensed or openly permitted test set, fixed model versions, repeatable measurements, and the underlying result records can be released together.

Technical references and accountability

The review categories are informed by music-source-separation research and public audio-processing documentation.

Reviewed and published by: MR Maker / MarineWave