Fix AI Vocal Fizz Without Rebuilding the Song: EditLab Developer Repair Walkthrough
Fix the Fizz. Keep the Vocal.
Can a finished AI-generated recording lose its buzzy vocal artifact without losing the voice that made the song worth keeping? Developer Mark Kaminsky provided a focused EditLab Tracker demonstration built around that question.
This walkthrough maps the actual example Mark described and supplied. It explains the producer decision, where to listen, and what remains to be independently verified.
Evidence standard: Developer-supplied, non-Suno audio examples from October 21, 2026. Not a JR hands-on benchmark, not an endorsed product test and not proof of performance on Suno v6.A keeper recording with one unwanted layer.
The song was made by Mark Kaminsky with AI assistance, not in Suno. Its lead vocal carried an unwanted high-frequency buzz or fizz. Mark's proposed repair was not another prompt or full-song regeneration. It was to use Tracker's separation workflow to isolate the unwanted component, mute that layer and keep the vocal.
He describes the isolated fizz as significant above 8 kHz, but the numbers and the audible result have not been independently verified by JR. The demonstration is useful because it makes the decision testable: preserve the source; remove only the troublesome part; then compare the reconstructed full mix against the original.
Import. Split. Mute. A/B.
Mark summarized the October demonstration as “drop the song in, split it, mute the fizz track, export.” The four stages below are an explanatory reconstruction of that description—not a screen-recorded JR session.
The updated splitter
On November 3, Mark reported that Tracker could now place bass, guitar, piano, vocal, drums and fizz on separate lanes. The developer also supplied two images of the newer version: the song-split screenshot and the split-window screenshot. These developer-supplied images are available as Shopify-hosted editorial screenshots; they do not independently verify the audio repair.
What to listen for: the buzz versus the repaired vocal.
Mark attached one MP3 containing three sequential comparison passages: (1) the vocal with its buzzy fizz, (2) the same four bars after the FIZZ lane was muted, and (3) the isolated fizz increased by 30 dB for analysis. This sequence makes the intervention understandable without falsely presenting three separately mastered tracks.
Example 1 — Original buzz, fizz muted, isolated fizz
Play Mark's original three-part demonstration in order: first the buzzy vocal, then the same four bars with the FIZZ lane muted, and finally the isolated fizz boosted by 30 dB. These are sequential passages in one MP3, not three different songs.
Listening checklist: Is the fizz less distracting? Does the repaired voice retain consonants and air? Do cymbals, consonant edges, room tone or transient details sound damaged? Is the difference still convincing at matched loudness? The best repair is not necessarily the most aggressive one.
The same vocal. A different bed.
Mark's second MP3 contrasts eight bars of the original song with the same eight bars after the vocal was retained and the music beneath it replaced. That is a replacement/arrangement example, not evidence of a fizz cleanup. It belongs beside the first comparison because it demonstrates another reason a separated project might be valuable.
Example 2 — Keep the vocal, change the arrangement
This single MP3 plays eight bars of the original recording, followed by the same vocal over new accompaniment. Listen for what changes musically, and what should remain recognizable in the performance.
Before
Eight bars of the original arrangement with the vocal and backing music together.
After
The same vocal over reworked accompaniment. Check whether the preserved vocal still fits harmonically, rhythmically and emotionally.
Useful possibility. Not a verified universal fix.
What is supported: Mark supplied examples and explained their signal path; the application screenshot is real; his developer claims support a plausible route for targeted edits without deliberate whole-song regeneration.
What remains unproven here: JR has not independently performed this specific repair in Tracker, measured the original and repaired files, compared the outputs with other tools, or established that Tracker will remove similar artifacts from Suno v6 songs. Some separation errors cannot be removed without changing or damaging desired material.
In November, Mark additionally offered to run a Suno v6 song with buried vocals through the newer splitter and send the original, split and result. That comparison is a sensible future teaching example, but it should be published only when the source is supplied, rights are cleared and the result is evaluated at comparable loudness.
Continue the connected learning path
Diagnose shimmer, hiss and generation artifacts →
Who made it, what was supplied, and what to verify.
Mark Kaminsky has described approximately 20 years working with tracker-based music applications and developing software for Apple platforms since 2010. He built Tracker to give himself more part-level control over AI-produced songs; he also records as Toyota X. These are his autobiographical statements, not independently established career records.
Primary evidence: Mark's October 21 emails supplying the two MP3s and the Tracker 1.5 screenshot; his November 2 and November 3 follow-ups on updated separation and his permission to reproduce the examples. The original MP3 recordings are hosted above with their filenames preserved in the download links. The composition video he provided separately is a creative-use illustration, not a repair test: see the separate developer video.
Commercial disclosure: This is independent educational coverage. JR has been offered complimentary Pro access for testing; no paid placement, affiliate deal or endorsement has been agreed. Software specifications, offers and version support should be confirmed directly with the developer.
Editorial note: Both original developer-provided MP3 demonstrations are included above for readers to inspect. This remains a documented developer case study, not an independent Jack Righteous Suno performance benchmark or guarantee that a given audio defect is repairable.
