Upgrade speech recognition models

Package.swift

@@ -12,15 +12,15 @@ let package = Package(
         .executable(name: "OpenTypeCLI", targets: ["OpenTypeCLI"]),
     ],
     dependencies: [
-        .package(url: "https://github.com/argmaxinc/WhisperKit.git", from: "1.0.0"),
+        .package(url: "https://github.com/argmaxinc/argmax-oss-swift.git", from: "1.0.0"),
         .package(url: "https://github.com/huggingface/swift-transformers", from: "1.3.3"),
         .package(url: "https://github.com/ml-explore/mlx-swift-lm", branch: "main"),
     ],
     targets: [
         .executableTarget(
             name: "OpenType",
             dependencies: [
-                .product(name: "WhisperKit", package: "WhisperKit"),
+                .product(name: "WhisperKit", package: "argmax-oss-swift"),
                 .product(name: "Hub", package: "swift-transformers"),
                 .product(name: "Tokenizers", package: "swift-transformers"),
                 .product(name: "MLXLLM", package: "mlx-swift-lm"),

README.md

@@ -15,7 +15,7 @@
 [![Apple Silicon](https://img.shields.io/badge/Apple%20Silicon-M1%2FM2%2FM3%2FM4-black?style=flat-square&logo=apple&logoColor=white)](https://www.apple.com/mac/m1/)
 [![License](https://img.shields.io/badge/license-MIT-blue?style=flat-square)](LICENSE)
 
-[![WhisperKit](https://img.shields.io/badge/Powered%20by-WhisperKit-blue?style=flat-square)](https://github.com/argmaxinc/WhisperKit)
+[![WhisperKit](https://img.shields.io/badge/Powered%20by-WhisperKit-blue?style=flat-square)](https://github.com/argmaxinc/argmax-oss-swift)
 [![MLX](https://img.shields.io/badge/Powered%20by-MLX--LM-orange?style=flat-square)](https://github.com/ml-explore/mlx-swift-lm)
 
 [Website](https://opentype.idevlab.dev) · [中文文档](README_zh.md)
@@ -161,7 +161,7 @@ scripts/
 
 ## Tech Stack
 
-- [WhisperKit](https://github.com/argmaxinc/WhisperKit) — offline Whisper speech recognition
+- [WhisperKit](https://github.com/argmaxinc/argmax-oss-swift) — offline Whisper speech recognition
 - [mlx-swift-lm](https://github.com/ml-explore/mlx-swift-lm) — local LLM inference on Apple Silicon (Qwen2.5 / Qwen3)
 - **SwiftUI + AppKit** — native macOS UI
 - **ScreenCaptureKit + Vision** — screen OCR

README_zh.md

@@ -15,7 +15,7 @@
 [![Apple Silicon](https://img.shields.io/badge/Apple%20Silicon-M1%2FM2%2FM3%2FM4-black?style=flat-square&logo=apple&logoColor=white)](https://www.apple.com/mac/m1/)
 [![License](https://img.shields.io/badge/license-MIT-blue?style=flat-square)](LICENSE)
 
-[![WhisperKit](https://img.shields.io/badge/驱动-WhisperKit-blue?style=flat-square)](https://github.com/argmaxinc/WhisperKit)
+[![WhisperKit](https://img.shields.io/badge/驱动-WhisperKit-blue?style=flat-square)](https://github.com/argmaxinc/argmax-oss-swift)
 [![MLX](https://img.shields.io/badge/驱动-MLX--LM-orange?style=flat-square)](https://github.com/ml-explore/mlx-swift-lm)
 
 [官网](https://opentype.idevlab.dev) · [English](README.md)
@@ -161,7 +161,7 @@ scripts/
 
 ## 技术栈
 
-- [WhisperKit](https://github.com/argmaxinc/WhisperKit) — 离线 Whisper 语音识别
+- [WhisperKit](https://github.com/argmaxinc/argmax-oss-swift) — 离线 Whisper 语音识别
 - [mlx-swift-lm](https://github.com/ml-explore/mlx-swift-lm) — Apple Silicon 本地 LLM 推理（Qwen2.5 / Qwen3）
 - **SwiftUI + AppKit** — macOS 原生 UI
 - **ScreenCaptureKit + Vision** — 屏幕 OCR

Sources/App/VoicePipeline+Models.swift

@@ -115,6 +115,17 @@ extension VoicePipeline {
         case .qwen3:
             let settings = appState.settings
             let catalog = ModelCatalog.shared
+            if !LocalASRRuntime.isReady(for: .qwen3),
+               !catalog.asrModelPath(for: settings.qwenASRModel).isEmpty {
+                do {
+                    _ = try await LocalASRRuntime.ensurePythonPath(
+                        for: .qwen3,
+                        preferredPath: settings.localASRPythonPath
+                    )
+                } catch {
+                    Log.error("[VoicePipeline] Qwen ASR runtime migration failed: \(error.localizedDescription)")
+                }
+            }
             guard localASRIsAvailable(settings.qwenASRModel) else {
                 qwenSpeechEngine = nil
                 markSpeechModelDownloadRequired(showInStatus: requestPermission)

Sources/Config/AppSettings.swift

@@ -357,7 +357,7 @@ final class AppSettings: ObservableObject {
         appIconAppearance = AppIconAppearance(rawValue: ud.string(forKey: Key.appIconAppearance.rawValue) ?? "") ?? .system
         volcAppKey = ud.string(forKey: Key.volcAppKey.rawValue) ?? ""
         volcAccessKey = ud.string(forKey: Key.volcAccessKey.rawValue) ?? ""
-        volcResourceId = ud.string(forKey: Key.volcResourceId.rawValue) ?? "volc.bigasr.sauc.duration"
+        volcResourceId = ud.string(forKey: Key.volcResourceId.rawValue) ?? VolcASRModel.recommended.rawValue
         localASRPythonPath = ud.string(forKey: Key.localASRPythonPath.rawValue) ?? LocalASRConfiguration.defaultPythonPath
         qwenASRModel = ud.string(forKey: Key.qwenASRModel.rawValue)
             ?? ud.string(forKey: Key.qwenASRModelPath.rawValue)

Sources/Config/VolcASRModel.swift

@@ -0,0 +1,17 @@
+import Foundation
+
+enum VolcASRModel: String, CaseIterable, Identifiable {
+    case seedASR2 = "volc.seedasr.sauc.duration"
+    case bigASR1 = "volc.bigasr.sauc.duration"
+
+    var id: String { rawValue }
+
+    var displayName: String {
+        switch self {
+        case .seedASR2: return L("volc.model.seed_asr_2")
+        case .bigASR1: return L("volc.model.big_asr_1")
+        }
+    }
+
+    static let recommended = VolcASRModel.seedASR2
+}

Sources/Resources/en.lproj/Localizable.strings

@@ -339,6 +339,9 @@
 
 /* ── Volc ASR ── */
 "volc.config_hint" = "Enter your Volcengine Doubao ASR credentials. Get them from the Volcengine console.";
+"volc.model" = "Recognition model";
+"volc.model.seed_asr_2" = "SeedASR 2.0 (Recommended)";
+"volc.model.big_asr_1" = "BigASR 1.0 (Legacy)";
 "volc.app_key" = "App Key (App ID)";
 "volc.access_key" = "Access Token";
 "volc.resource_id" = "Resource ID";

Sources/Resources/zh-Hans.lproj/Localizable.strings

@@ -339,6 +339,9 @@
 
 /* ── 豆包语音识别 ── */
 "volc.config_hint" = "填写火山引擎豆包语音识别凭据，可从火山引擎控制台获取。";
+"volc.model" = "识别模型";
+"volc.model.seed_asr_2" = "SeedASR 2.0（推荐）";
+"volc.model.big_asr_1" = "BigASR 1.0（旧版）";
 "volc.app_key" = "App Key（应用 ID）";
 "volc.access_key" = "Access Token";
 "volc.resource_id" = "Resource ID（资源 ID）";

Sources/Speech/AppleSpeechAnalyzer.swift

@@ -0,0 +1,79 @@
+import AVFoundation
+import Foundation
+@preconcurrency import Speech
+
+enum AppleSpeechAnalyzer {
+    static func prepare(locale: Locale) async throws {
+        let transcriber = try await makeTranscriber(locale: locale)
+        try await ensureModel(for: transcriber)
+    }
+
+    static func transcribe(audioURL: URL, locale: Locale) async throws -> String {
+        let transcriber = try await makeTranscriber(locale: locale)
+        try await ensureModel(for: transcriber)
+
+        let file = try AVAudioFile(forReading: audioURL)
+        let resultTask = Task<String, Error> {
+            var transcript = ""
+            for try await result in transcriber.results where result.isFinal {
+                transcript += String(result.text.characters)
+            }
+            return transcript
+        }
+
+        let analyzer = SpeechAnalyzer(
+            modules: [transcriber],
+            options: .init(priority: .userInitiated, modelRetention: .lingering)
+        )
+        do {
+            if let lastSample = try await analyzer.analyzeSequence(from: file) {
+                try await analyzer.finalizeAndFinish(through: lastSample)
+            } else {
+                await analyzer.cancelAndFinishNow()
+            }
+            return try await resultTask.value
+        } catch {
+            resultTask.cancel()
+            await analyzer.cancelAndFinishNow()
+            throw error
+        }
+    }
+
+    private static func makeTranscriber(locale: Locale) async throws -> SpeechTranscriber {
+        guard SpeechTranscriber.isAvailable else {
+            throw AppleSpeechAnalyzerError.unavailable
+        }
+        guard let supported = await SpeechTranscriber.supportedLocale(equivalentTo: locale) else {
+            throw AppleSpeechAnalyzerError.unsupportedLocale(locale.identifier)
+        }
+        return SpeechTranscriber(locale: supported, preset: .transcription)
+    }
+
+    private static func ensureModel(for transcriber: SpeechTranscriber) async throws {
+        let modules: [any SpeechModule] = [transcriber]
+        if await AssetInventory.status(forModules: modules) == .installed { return }
+        if let request = try await AssetInventory.assetInstallationRequest(supporting: modules) {
+            try await request.downloadAndInstall()
+        }
+        guard await AssetInventory.status(forModules: modules) == .installed else {
+            throw AppleSpeechAnalyzerError.modelUnavailable
+        }
+    }
+}
+
+enum AppleSpeechAnalyzerError: LocalizedError {
+    case unavailable
+    case unsupportedLocale(String)
+    case modelUnavailable
+
+    var errorDescription: String? {
+        switch self {
+        case .unavailable:
+            return "SpeechAnalyzer is unavailable on this Mac"
+        case .unsupportedLocale(let locale):
+            return "SpeechAnalyzer does not support locale \(locale)"
+        case .modelUnavailable:
+            return "SpeechAnalyzer model is not installed"
+        }
+    }
+}

Sources/Speech/AppleSpeechEngine.swift

@@ -2,7 +2,7 @@ import Foundation
 @preconcurrency import Speech
 import AVFoundation
 
-final class AppleSpeechEngine: SpeechEngine, @unchecked Sendable {
+final class LegacyAppleSpeechEngine: SpeechEngine, @unchecked Sendable {
     private var recognizer: SFSpeechRecognizer
     private(set) var isReady = false
     private var recognitionTask: SFSpeechRecognitionTask?

Sources/Speech/AppleSpeechEngineAdapter.swift

@@ -0,0 +1,88 @@
+import AVFoundation
+import Foundation
+
+final class AppleSpeechEngine: SpeechEngine, @unchecked Sendable {
+    private let locale: Locale
+    private let legacy: LegacyAppleSpeechEngine
+
+    init(locale: Locale = Locale(identifier: "zh-CN")) {
+        self.locale = locale
+        self.legacy = LegacyAppleSpeechEngine(locale: locale)
+    }
+
+    var isReady: Bool { legacy.isReady }
+    var supportsStreaming: Bool { true }
+
+    func requestAccess() {
+        legacy.requestAccess()
+    }
+
+    func prepare() async {
+        do {
+            try await AppleSpeechAnalyzer.prepare(locale: locale)
+        } catch {
+            Log.info("[AppleSpeech] SpeechAnalyzer preparation deferred: \(error.localizedDescription)")
+        }
+    }
+
+    func startListening(language: String?, onPartialResult: @escaping @Sendable (String) -> Void) {
+        legacy.startListening(language: language, onPartialResult: onPartialResult)
+    }
+
+    func appendAudioBuffer(_ buffer: AVAudioPCMBuffer) {
+        legacy.appendAudioBuffer(buffer)
+    }
+
+    func finishListening(audioURL: URL?, language: String?) async throws -> String {
+        let fallbackTask = Task {
+            try await legacy.finishListening(audioURL: audioURL, language: language)
+        }
+        guard let audioURL else { return try await fallbackTask.value }
+
+        do {
+            let text = try await AppleSpeechAnalyzer.transcribe(
+                audioURL: audioURL,
+                locale: resolvedLocale(for: language)
+            )
+            legacy.cancelListening()
+            let fallback = try? await fallbackTask.value
+            return text.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty
+                ? fallback ?? ""
+                : text
+        } catch {
+            Log.info("[AppleSpeech] SpeechAnalyzer failed, using legacy result: \(error.localizedDescription)")
+            return try await fallbackTask.value
+        }
+    }
+
+    func cancelListening() {
+        legacy.cancelListening()
+    }
+
+    func transcribe(audioURL: URL?, language: String?) async throws -> String {
+        guard let audioURL else { throw AppleSpeechError.noAudioFile }
+        do {
+            let text = try await AppleSpeechAnalyzer.transcribe(
+                audioURL: audioURL,
+                locale: resolvedLocale(for: language)
+            )
+            if !text.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty {
+                return text
+            }
+        } catch {
+            Log.info("[AppleSpeech] SpeechAnalyzer failed, using legacy recognizer: \(error.localizedDescription)")
+        }
+        return try await legacy.transcribe(audioURL: audioURL, language: language)
+    }
+
+    private func resolvedLocale(for language: String?) -> Locale {
+        switch language {
+        case "zh": return Locale(identifier: "zh-CN")
+        case "en": return Locale(identifier: "en-US")
+        case "ja": return Locale(identifier: "ja-JP")
+        case "ko": return Locale(identifier: "ko-KR")
+        case "yue": return Locale(identifier: "yue-CN")
+        default: return locale
+        }
+    }
+}