Skip to content

Transcribe

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

Transcribe

README.md

Transcribe — on-device speech-to-text, any ASR model in the catalog

The speech-to-text model runner: one app for every asr catalog entry — Whisper large-v3-turbo (100 languages), Qwen3-ASR 1.7B, Parakeet-TDT 0.6B v3. Record a clip, choose a file, or use the bundled demo; pick the model in the picker; transcription runs fully on device (first use downloads the model, then it loads from the local cache).

let transcriber = try await KitTranscriber(catalog: "whisper-large-v3-turbo")
let samples = try AudioFile.pcm16kMono(url)  // any wav/m4a/mp3 → 16 kHz mono Float
let result = try await transcriber.transcribe(samples: samples)
// result.text, result.language

Run it

GUI (macOS / iOS):

open Transcribe.xcodeproj   # → Run, pick a model in the picker

CLI (macOS, headless — agents verify with this):

swift run transcribe-cli --audio sample.wav --model whisper-large-v3-turbo
swift run transcribe-cli --list-models

The transcript goes to stdout; download progress and the detected language go to stderr.

The take-home: Sources/QuickStart.swift

One typed function, no UI imports — audio file in, Transcription out:

func transcribe(audio url: URL, model id: String, …) async throws -> Transcription

Both shells call exactly this function: the GUI view model is a display shell over it, the CLI an argument shell. Want transcription in your own app? Copy that file — the glue (file → 16 kHz mono PCM, mic capture, id → engine dispatch) is all kit API (AudioFile.pcm16kMono, MicRecorder, KitTranscriber), so it runs as pasted.

The function loads the model per call — the honest end-to-end shape. If your app transcribes repeatedly, hold the KitTranscriber and call transcribe(samples:) on it per clip.

Integration checklist

  • SPM: github.com/john-rocky/coreai-kit → product CoreAIKit
  • Info.plist: NSMicrophoneUsageDescription (only if you record; the permission prompt and record button are app chrome — MicRecorder does the capture)
  • Entitlements (iOS): com.apple.developer.kernel.increased-memory-limit for Whisper (1.6 GB fp16 graph; 3.2 GB AOT bundle on iPhone)
  • First run downloads the model → cached in Application Support (progress callback)
  • Measure in Release — Debug is ~3× slower on per-token host work

Models

Catalog id Model Platforms Notes
whisper-large-v3-turbo Whisper large-v3-turbo macOS + iOS 100 languages, auto-detect, ≤30 s window
qwen3-asr-1.7b Qwen3-ASR 1.7B macOS LLM-engine ASR, 52 languages
parakeet-tdt-0.6b-v3 Parakeet-TDT 0.6B v3 macOS TDT transducer, fastest of the three