feat: add WhisperKit transcriber and transcribe probe

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

Sources/SstProbe/TranscribeProbe.swift

@@ -0,0 +1,17 @@
+import Foundation
+import SstWhisper
+
+func runTranscribeProbe() async {
+    guard let samples = await recordSamples(seconds: 6) else { return }
+    print("Model laden (eerste keer: download ≈1.5 GB)…")
+    let transcriber = WhisperKitTranscriber()
+    do {
+        try await transcriber.load()
+        let started = Date.now
+        let text = try await transcriber.transcribe(samples)
+        let elapsed = Date.now.timeIntervalSince(started)
+        print(String(format: "(%.2f s) ", elapsed) + text)
+    } catch {
+        print("mislukt:", error.localizedDescription)
+    }
+}

Sources/SstProbe/main.swift

@@ -6,6 +6,8 @@ case "hotkey":
     runHotkeyProbe()
 case "record":
     await runRecordProbe()
+case "transcribe":
+    await runTranscribeProbe()
 default:
-    print("usage: sst-probe hotkey|record")
+    print("usage: sst-probe hotkey|record|transcribe")
 }

Sources/SstWhisper/WhisperKitTranscriber.swift

@@ -0,0 +1,40 @@
+import Foundation
+import SstCore
+import WhisperKit
+
+public final class WhisperKitTranscriber: Transcriber {
+    public enum TranscriberError: Error, LocalizedError {
+        case notLoaded
+
+        public var errorDescription: String? { "Spraakmodel is nog niet geladen." }
+    }
+
+    private let modelName: String
+    private var pipeline: WhisperKit?
+
+    public init(model: String = "large-v3-v20240930_turbo") {
+        self.modelName = model
+    }
+
+    /// Downloads the model on first run (≈1.5 GB) and loads it into memory.
+    public func load() async throws {
+        pipeline = try await WhisperKit(WhisperKitConfig(model: modelName))
+    }
+
+    public func transcribe(_ samples: [Float]) async throws -> String {
+        guard let pipeline else { throw TranscriberError.notLoaded }
+        // language nil + detectLanguage: Whisper picks NL or EN per utterance.
+        let options = DecodingOptions(
+            task: .transcribe,
+            language: nil,
+            detectLanguage: true,
+            skipSpecialTokens: true,
+            withoutTimestamps: true
+        )
+        let results = try await pipeline.transcribe(audioArray: samples, decodeOptions: options)
+        return results
+            .map(\.text)
+            .joined(separator: " ")
+            .trimmingCharacters(in: .whitespacesAndNewlines)
+    }
+}