Introducing StemStash: automatic stem organization for producers
· Dave Nunes
StemStash is live. It’s a tool for producers and mix engineers that takes a folder of exported stems and returns them classified by instrument type, tagged with BPM and key, renamed, and sorted into consistent group folders, ready to drop straight back into a DAW.
How do I organize exported stems automatically?
Every uploaded stem is assigned to one of five groups: DRUMS, BASS, MELODY, VOCALS, or FX. If a project includes a full mix or master bounce, StemStash detects that file separately and keeps it at the top level of the export instead of filing it into a group folder.
Files are renamed using the detected group, sound type, project BPM, and project name. A file that came in as kick (3) FINAL_v2.wav goes out as something like DRUMS_KICK_DRUM_140_MIDNIGHT_DRIVE.wav, and a master bounce goes out with its own format, for example MIDNIGHT_DRIVE - MASTER - 140 BPM - Amin.wav.
A single project supports up to 500 stems and 20 GB combined, across WAV, AIFF, FLAC, and MP3, with resumable uploads so a large batch doesn’t have to restart from zero if a connection drops.
How does StemStash detect BPM, key, and instrument type?
Three separate detection passes run on every project, and none of them are a generative AI call.
- Instrument type comes from CLAP (laion/clap-htsat-unfused), a zero-shot audio classification model, matched against dozens of candidate sound categories using the full file rather than a short clip. The same stem gets the same label every time it’s analyzed.
- Tempo comes from onset-based beat tracking, read from the master file when a project has one, or cross-checked against the drum and bass stems when it doesn’t.
- Key comes from chroma pitch-class correlation, matching each file’s pitch content against major and minor key profiles, with drum, FX, and silent stems left out of the calculation so they can’t skew the result.
Silence is flagged by peak level across the whole file, not a short sample of it, and master-bounce detection weighs filename, classification confidence, and duration together rather than just picking the loudest file. Every value ships with a visible confidence score and can be corrected in a review table before anything exports.
Is StemStash an AI stem naming tool?
Partly, and the distinction matters. Instrument classification uses a machine learning model, CLAP, run zero-shot and not fine-tuned on any user’s uploads, so that one step genuinely is AI. Tempo and key detection are not: those are established signal-processing techniques, onset-based beat tracking and chroma pitch-class correlation, with no model involved. None of it is generative AI. Filenames are built from a fixed template, group, sound type, BPM, project name, not generated as freeform text, and nothing about the audio itself is created, rewritten, or altered.
That’s also why the results are repeatable. Run the same stems through StemStash twice and the classification, BPM, and key come back identical both times, since there’s no randomness or per-request model variance in the pipeline. The CLAP model itself was trained on a public dataset of more than 630,000 audio samples before StemStash ever used it. It isn’t trained or retrained on anything a user uploads.
Is my audio used to train AI models?
No. Uploaded stems are never used to train any AI model, by StemStash or by anyone else, and they’re never inspected, downloaded, or listened to by a person as part of normal operation. Files sit in a private, per-account storage area with no public or authenticated browsing access attached to it. The only way a file is ever retrieved is through a short-lived, single-use signed link, and it expires within a minute. Everything is auto-deleted no later than two hours after export completes, or immediately if you delete it yourself from the export screen.
Payments are handled by Stripe, which is PCI-DSS compliant, and StemStash never sees or stores a full card number. StemStash is currently rated 4.8 out of 5 on Trustpilot, based on 243 verified reviews.
How do I get started?
The free Hobby tier includes 10 credits to try StemStash with, no card required. Get started, drop in the stems from your next export, review the detected groups, filenames, BPM, and key, correct anything that needs it, and export a folder that’s already organized.