ScribeKitInstallation
Add ScribeKit to a Swift package or an Xcode project.
Swift Package Manager
Add the package to Package.swift:
dependencies: [
.package(url: "https://github.com/pieralukasz/ScribeKit.git", from: "0.1.0"),
],
targets: [
.target(name: "MyApp", dependencies: ["ScribeKit"]),
]In Xcode, use File → Add Package Dependencies… and paste https://github.com/pieralukasz/ScribeKit.
ScribeKit pins FluidAudio to an exact version, so an update of FluidAudio cannot change your results without a new ScribeKit release.
Models
Nothing is bundled. The first call that needs a model downloads it from Hugging Face and caches it in ~/Library/Application Support/FluidAudio/Models:
| Model | Folder | Size |
|---|---|---|
| Parakeet TDT 0.6B v3 (speech) | parakeet-tdt-0.6b-v3 | about 460 MB |
| pyannote community-1 (speakers) | speaker-diarization | about 20 MB |
Later runs load from the cache and need no network. Load them ahead of time with prepare, so the first transcription does not wait. See Models.
Try it from the terminal
The package ships a small scribe command:
git clone https://github.com/pieralukasz/ScribeKit.git
cd ScribeKit
swift run -c release scribe prepare --diarize
swift run -c release scribe meeting.m4a --lang enSee Command line for every option.
Licenses
ScribeKit is MIT licensed. The models are not part of it: Parakeet is released by NVIDIA under CC-BY-4.0, the diarization model comes from pyannote, and both are converted to Core ML by FluidInference. Check their licenses before you ship.