demorphus
Our internal tests now exceed 2,500 tracks, with an overall 98% Human Made success rate after cleaning. The September 29, 2026 overview covers Suno, Udio, Mureka and Mubert, with more than 450 tracks tested per generator.
Now tested with Suno v6
Demorphus Cleaner has now been tested with Suno v6. This coverage update was added October 30, 2026.
The 2,500+ track overview below is dated September 29. No separate Suno v6 sample count or success rate is published here; the overall 98% figure should not be read as a v6-specific result.
Overall results: 98% success rate
Across more than 2,500 tracks submitted to Cleaner, the overall success rate is 98%. Success here means receiving a Human Made label after processing. This percentage does not measure audio fidelity or streaming-quality certification.
This overview is updated manually. The figures replace the previous milestone of 1,289 successful outcomes; they are not added to it. The 2,500+ figure counts tested tracks, not just those receiving Human Made labels.
More than 450 tracks per generator
| Generator | Tracks tested | Observed outcome |
|---|---|---|
| Suno | More than 450 | Human Made achieved |
| Udio | More than 450 | Human Made achieved |
| Mureka | More than 450 | Human Made achieved |
| Mubert | More than 450 | Human Made achieved |
The 98% figure applies to the overall trials, not a separate rate for each row. Exact per-generator counts and corresponding failures are not detailed here. This table shows coverage without inventing a breakdown of the total.
Original WAV delivers better results
The weekend trials show better results with WAV files. For uploads, prefer the original WAV downloaded from your generator when available. MP3 and FLAC are still accepted.
Converting an MP3 to WAV does not recover detail lost during compression. Keep the best available source and compare the output through listening.
Method and available data
Source: Demorphus internal test overview supplied on September 29, 2026. This is not an independent study. Attempts per track and intermediate scores are not detailed. The 98% figure therefore does not measure first-attempt success.
Individual files, reports, model versions and detailed listening assessments are not published here. A new track or detector version may produce a different outcome.
Check your output
Keep the original and compare an exposed vocal, dense chorus and transition at similar playback levels. A Human Made verdict does not replace listening or change a track’s origin.
To track your trials, record source format, recipe, processing attempts, score and listening notes. See our audio comparison guide.
From detection results to a release
These trials measure the detector response to processed files. They do not measure an acceptance rate across distributors. Keep your original, selected export and dated report together, then follow the release preparation workflow for the submission itself.
Platform uploads: our release experience
We have uploaded thousands of processed tracks to streaming platforms without an AI flag observed. This release experience was added to our overview on October 30, 2026.
A platform-by-platform count, observation period and flag-checking method are not published here. This is an internal observation, not an independently audited platform benchmark. It does not establish access to recommendations or measured royalty income.
These uploads are reported separately from the September 29 detector trials. We do not add the counts together or apply the 98% detector result to platform releases. See what each release outcome measures.
Prepared with AI assistance. Sources below.