Skip to content
ToneTrace

Features

What ToneTrace does today

Everything on this page is in the current build. Nothing here is a roadmap item, and the app does all of it on your own computer.

Drop in a folder, get a library

Drag files or folders onto ToneTrace, or pick a folder from the native dialog. Every audio file inside is queued and analysed in the background while you keep browsing.

  • Files stay where they are. ToneTrace never copies, moves or renames anything.
  • Live counts in the sidebar: analysed, pending, analysing, failed.
  • Formats: MP3, WAV, FLAC, M4A, AAC, OGG, WMA and AIFF.
For example

Drop a folder of 400 drum one-shots. At about a second and a half per file on a laptop CPU, they are all tagged in roughly ten minutes, and you can start searching the first ones straight away.

Tags that say how sure they are

Two models listen to every file. YAMNet names instruments and sound events. CLAP describes mood, genre and production character. Each tag shows the model's own confidence, not an invented score.

  • Instrument tags come from Google's YAMNet, trained on AudioSet.
  • Mood, genre and character tags come from LAION's CLAP, matched against a curated vocabulary of descriptions.
  • When YAMNet is torn between similar instruments, CLAP's opinion is weighed in to settle it. Neither model can overrule the other on its own.
For example

A strummed acoustic chord comes back as Guitar, with blues and funk from the mood model. A solo ukulele recording comes back as Ukulele. A voice memo comes back as Speech with dry vocal.

See it, hear it, loop it

Every card shows the real waveform of the file, drawn from the same decode the models listened to. Click to play, click the waveform to seek, loop it, set the volume.

  • One player for the whole app, so you never end up with two things playing.
  • Seeking works on any format the app imports.
  • Duration and file size are shown on every card and in the detail view.
For example

Hover a card, hit play, and the transport bar at the bottom picks it up with a scrubber, loop toggle and volume. Open the file for a full-width waveform you can click through.

Your folders, as they are

ToneTrace builds a folder tree from the paths in your library. Click a folder to filter the grid, include or exclude subfolders, and the tree updates as you add more.

  • The library is remembered between launches and loads on start.
  • Files that have gone missing are skipped rather than shown as broken cards.
  • Remove files from the library without touching them on disk.
For example

Import your Drums, Bass and Synths folders and the sidebar shows them nested the way they are on disk, with counts, and the grid follows whichever one you select.

Never analysed twice

Every file is fingerprinted with a content hash before anything else happens. If ToneTrace has heard that exact audio before, the tags come straight from the local database.

  • The hash is of the audio file's bytes, so moved and renamed files are recognised.
  • Edited or re-exported files have new content and are analysed fresh.
  • Analysis results are versioned, so an improved pipeline can re-tag on request rather than silently mixing old and new.
For example

Reorganise your sample folders, rename half of them, then re-import. Every file that was already analysed is recognised by its content and appears instantly.

Pick a sound, get its neighbours

Select any file and choose Find Similar. ToneTrace can rank your library by audio content, using the fingerprint the language-audio model produced, or by shared tags.

  • By sound: how close each file's audio fingerprint sits to the one you picked, computed locally.
  • By tags: overlap of tags, weighted by confidence.
  • Results reuse the same cards, so you can play, select and drag from them directly.
For example

Select a snare you like and ask for similar by sound. The rest of your library is ordered by how close each file's audio fingerprint sits to it, with a match figure on every card.

Tags your other tools can read

Select files and choose Export tags. ToneTrace writes the tags into the files using standard metadata, so DAW browsers and sample managers that read it will show them.

  • MP3: ID3 tags (genre, comment and a ToneTrace frame).
  • WAV: Broadcast Wave bext chunk plus LIST/INFO keywords and comment.
  • Everything else: a JSON sidecar next to the file.
  • By default, up to ten tags at or above 30 % confidence are written.
For example

Export a folder of WAV one-shots and each file gains a Broadcast Wave description and INFO keywords. The audio data is untouched, byte for byte, and exporting twice does not grow the file.

Drag straight into your session

Every card is a native file drag. Drop it into your DAW, a sampler, or a Finder or Explorer window exactly as you would from the desktop. No plugin, no bridge.

  • Works with any application that accepts file drops from the operating system.
  • Multi-select drags the whole selection.
  • Nothing is converted or copied on the way; the DAW gets the original file.
For example

Find the hi-hat, hold the card, drag it onto a drum rack. Select five and drag them all at once. The transport bar's Send to DAW button drags whatever is playing.

Runs where you work

ToneTrace is a desktop app built with Electron and React, with builds for macOS, Windows and Linux. Both models ship inside the app and run locally by default.

  • macOS on Apple Silicon and Intel, Windows, and Linux.
  • No sign-in and no telemetry in the current build.
  • Installers are being finalised; early-access testers get builds first.
For example

Install it, point it at a folder, and the current build works with the network cable unplugged.

Try it on your own library

Early access is free. Pricing will be announced at launch.

Request early access