AudioShake's stem tool is built for catalogs, not consumers

The tool moment

AudioShake is a company that sells source-separation technology, not a plugin a musician installs at home. Its own site states the technology can split music into its instrument stems and separate dialogue, music, and effects from a mixed recording, paired with a lyric-transcription tool that produces word-by-word synced lyrics. A product page lists the intended uses as remixing, mastering, karaoke, sync licensing, spatial mixes, transcription, and archival projects, and names its target customers as labels, publishers, producers, DJs, engineers, and app builders. Both pages list major rights holders among AudioShake's clients, including record labels, a sports broadcaster, and film and television studios, and describe the technology used on catalog recordings by artists including the Jackson 5, Nina Simone, and Whitney Houston. This reflects the site as retrieved 16 September 2026; it is a living listing, not a dated launch.

What the documents show

AudioShake's own materials frame the technology around a catalog's or rights holder's problem, not a bedroom remixer's: mastering a legacy recording with no surviving multitrack session, clearing a sync placement quickly, or making an old recording accessible through captioning. That framing is a business-positioning choice stated by the company, not an independently verified accuracy claim. The pages describe the vendor's own view of its output as 'state-of-the-art AI vocal isolation with best-in-class stem quality' that 'preserve[s] the integrity of the original performance,' language that is AudioShake's own characterization rather than a benchmark score or a named third-party test. The client list is likewise the company's own disclosure, not a verified case study.

What stays with the musician

Separating stems from a legacy recording with no surviving session tapes does not, by itself, settle who is allowed to authorize a new mix, a sync placement, or a reissue built from those stems. That is a rights-holder decision sitting above the technical separation step, and AudioShake's site addresses licensing and catalog use as a service context, not a legal opinion about any specific recording. A musician or estate considering a stem-based reissue still has to confirm who controls the underlying rights before the stems go near a release.

Judge it by listening

None of AudioShake's pages cite a specific accuracy percentage or a documented listening-test method, so there is no vendor benchmark to check by ear against a fixed number. Editorially, the fairest way to evaluate a claim about preserving a performance's integrity is to compare a separated stem against the original mix where instruments overlap most, listening for whether the isolated part still carries the recording's transients and room sound, or has been smoothed into something that only resembles the source.

• Is this recording being separated for internal catalog work, or for a public release needing its own rights clearance?
• Does the company state a specific accuracy figure, or only general language about being state of the art?
• Who currently controls the rights to the original recording the stems were pulled from?

A stem-separation vendor built for rights holders still leaves the rights question exactly where it was; the technology changes what can be extracted from a mix, not who is allowed to use it.