Go Meow Toolkit command guide

The current release is source-linked software for macOS/Linux. It requires Node.js 22 or newer, ffmpeg and ffprobe. From an authorized source checkout of gomeow.media, install its dependencies, then link the executable. Keep the checkout and dependencies in place. A public standalone installer is not available yet.

npm ci
npm link
gomeow-toolkit --help
gomeow-toolkit --version

After installation, run gomeow-toolkit from any directory. The commands below operate on local files, make no paid generation requests and preserve source audio. They do not require a Studio sign-in.

gomeow-toolkit inspect "recording.wav"
gomeow-toolkit duplicates "original.wav" "copy.wav"
gomeow-toolkit compare "original.wav" "tagged.wav"
gomeow-toolkit align "original.wav" "shifted.wav"
gomeow-toolkit match "excerpt.wav" "candidate.mp3"
gomeow-toolkit report "recording.wav" > observations.json

JSON goes to stdout; errors go to stderr. Exit 0 means the requested analysis completed, not that files match or meet a quality standard. Inspect/report return exit 2 for partial, undefined or failed measurements. Read each check state and the reported scope. Never redirect a report over the input file.

Decoded equality uses the selected first audio stream and recorded decoder version. Resampling, gain changes and compression can produce unequal decoded samples even for related recordings. Exact file hashes identify bytes, not songs or speakers.

Alignment reports positive offsets when corresponding audio occurs later in the second file. Difference measurements retain gain and polarity, cover only overlapping 8 kHz mono audio, and can miss stereo or high-frequency changes. Weak or competing evidence returns ambiguous.

Fingerprint results are candidate-match, ambiguous, no-match or insufficient-evidence. The method uses FFT band-energy sketches and records its thresholds. Tests cover synthetic varied audio, gain changes, excerpts, compression, resampling and negative cases. Broad field-audio accuracy, pitch changes and time stretching remain unevaluated; scores are not calibrated probabilities.

A forensic observation report does not prove authenticity, authorship, absence of editing or chain of custody. Its checksum is not a signature: someone can rewrite both report and checksum. Keep a separately trusted copy if you need to detect later changes.

Self-contained WAV, FLAC, MP3, Ogg and AIFF are supported. Playlists, network media references and other containers are rejected. Sources are limited to 2 GiB. Analysis supports at most eight channels and 192 kHz; ceilings are not an enterprise performance guarantee.

Inspect/report allow up to two hours. Decoded comparison, alignment and fingerprint matching currently require whole streams of at most 120 seconds. Canonical decoded output is bounded at 256 MiB, so high-rate multichannel material can hit the limit sooner. Additional streams are not analyzed.

gomeow-toolkit inspect "recording.wav" --seconds 30 --timeout-ms 60000

Only inspect accepts --seconds; prefix coverage never counts as whole-file completion. All commands accept a positive --timeout-ms up to 600000, with a five-minute default. Ctrl-C cancels the job and its decoder processes. One job runs per process. Sources are hashed and checked before and after processing; changed-source results are discarded.

Current local CLI commands have no per-use credit charge or Studio authentication gate. There is no paid Toolkit edition or checkout yet. The planned free edition should complete useful local jobs and export clean results. Batch workflows, deeper restoration and team administration are roadmap items, not existing paid unlocks.

Desktop workbench integration, Windows cancellation support, repair tools, persistent fingerprint libraries, signed installers and enterprise controls remain in development or planning. The Toolkit does not add new MusicScript language declarations.

• Go Meow Toolkit overview
• MusicScript language and CLI reference
• Go Meow Server and live sessions