@wovin/tranz 0.1.11 → 0.1.15

package/README.md

@@ -49,7 +49,7 @@ const transcriber = createMistralTranscriber({
 const result = await transcriber.transcribe({
   audioPath: './interview.mp3',
   diarize: true,
-  timestamps: 'word'
+  timestamps: 'segment'
 })
 
 console.log(result.text)
@@ -137,7 +137,7 @@ const result = await mistral.transcribe({
   apiKey: process.env.MISTRAL_API_KEY,
   model: 'voxtral-mini-latest',
   diarize: true,
-  timestampGranularity: 'word'
+  timestampGranularity: 'segment'
 })
 
 // Whisper provider (local)
@@ -264,7 +264,7 @@ Options for the `transcribe()` method:
 - `language?: string` - Language code (e.g., 'en', 'fr') - disables word timestamps
 - `model?: string` - Override default model
 - `diarize?: boolean` - Enable speaker diarization (default: true)
-- `timestamps?: 'word' | 'segment'` - Timestamp granularity (default: 'word')
+- `timestamps?: 'word' | 'segment'` - Timestamp granularity (default: 'segment' when diarize is true, disabled if language is set)
 - `autoSplit?: boolean` - Auto-split long audio (default: true)
 - `splitOutputDir?: string` - Directory for split segments (default: system temp)
 - `logger?: TranscribeLogger` - Custom logger

package/dist/index.d.ts

@@ -5,5 +5,6 @@ export { createProvider, MistralProvider, WhisperProvider, GreenPTProvider, VOXT
 export { autoSplitAudio, analyzeSplitPoints, detectSilenceRegions, getAudioDuration, findOptimalSplitPoints, splitAudioAtPoints, DEFAULT_SPLIT_CONFIG, type SplitConfig, type SilenceRegion, type SplitPoint, type AudioSegment, type SplitAnalysis, } from './utils/audio/split.ts';
 export { mergeTranscriptionResults, formatMergedText, type MergedTranscriptionResult, type WordData, } from './utils/audio/merge-results.ts';
 export { formatTranscriptWithPauses } from './utils/transcription/format.ts';
+export { detectAudioMimeType } from './utils/transcription/mime-detection.ts';
 export { createMistralTranscriber, transcribe, type TranscribeOptions, type MistralTranscriberConfig, type MistralTranscriber, } from './utils/transcription/transcribe.ts';
 //# sourceMappingURL=index.d.ts.map

package/dist/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA;;GAEG;AAGH,OAAO,EACL,cAAc,EACd,eAAe,EACf,eAAe,EACf,eAAe,EACf,cAAc,EACd,KAAK,YAAY,EACjB,KAAK,gBAAgB,EACrB,KAAK,mBAAmB,EACxB,KAAK,qBAAqB,GAC3B,MAAM,oCAAoC,CAAA;AAG3C,OAAO,EACL,cAAc,EACd,kBAAkB,EAClB,oBAAoB,EACpB,gBAAgB,EAChB,sBAAsB,EACtB,kBAAkB,EAClB,oBAAoB,EACpB,KAAK,WAAW,EAChB,KAAK,aAAa,EAClB,KAAK,UAAU,EACf,KAAK,YAAY,EACjB,KAAK,aAAa,GACnB,MAAM,wBAAwB,CAAA;AAG/B,OAAO,EACL,yBAAyB,EACzB,gBAAgB,EAChB,KAAK,yBAAyB,EAC9B,KAAK,QAAQ,GACd,MAAM,gCAAgC,CAAA;AAGvC,OAAO,EAAE,0BAA0B,EAAE,MAAM,iCAAiC,CAAA;AAG5E,OAAO,EACL,wBAAwB,EACxB,UAAU,EACV,KAAK,iBAAiB,EACtB,KAAK,wBAAwB,EAC7B,KAAK,kBAAkB,GACxB,MAAM,qCAAqC,CAAA"}
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA;;GAEG;AAGH,OAAO,EACL,cAAc,EACd,eAAe,EACf,eAAe,EACf,eAAe,EACf,cAAc,EACd,KAAK,YAAY,EACjB,KAAK,gBAAgB,EACrB,KAAK,mBAAmB,EACxB,KAAK,qBAAqB,GAC3B,MAAM,oCAAoC,CAAA;AAG3C,OAAO,EACL,cAAc,EACd,kBAAkB,EAClB,oBAAoB,EACpB,gBAAgB,EAChB,sBAAsB,EACtB,kBAAkB,EAClB,oBAAoB,EACpB,KAAK,WAAW,EAChB,KAAK,aAAa,EAClB,KAAK,UAAU,EACf,KAAK,YAAY,EACjB,KAAK,aAAa,GACnB,MAAM,wBAAwB,CAAA;AAG/B,OAAO,EACL,yBAAyB,EACzB,gBAAgB,EAChB,KAAK,yBAAyB,EAC9B,KAAK,QAAQ,GACd,MAAM,gCAAgC,CAAA;AAGvC,OAAO,EAAE,0BAA0B,EAAE,MAAM,iCAAiC,CAAA;AAG5E,OAAO,EAAE,mBAAmB,EAAE,MAAM,yCAAyC,CAAA;AAG7E,OAAO,EACL,wBAAwB,EACxB,UAAU,EACV,KAAK,iBAAiB,EACtB,KAAK,wBAAwB,EAC7B,KAAK,kBAAkB,GACxB,MAAM,qCAAqC,CAAA"}

package/dist/index.min.js

@@ -1,3 +1,7 @@
+var __defProp = Object.defineProperty;
+var __defNormalProp = (obj, key, value) => key in obj ? __defProp(obj, key, { enumerable: true, configurable: true, writable: true, value }) : obj[key] = value;
+var __publicField = (obj, key, value) => __defNormalProp(obj, typeof key !== "symbol" ? key + "" : key, value);
+
 // src/utils/transcription/providers.ts
 import { spawn } from "child_process";
 import * as fs from "fs";
@@ -57,26 +61,10 @@ function createProvider(providerName, config) {
       throw new Error(`Unknown provider: ${providerName}`);
   }
 }
-var WhisperProvider = class _WhisperProvider {
-  name = "whisper";
-  cacheDir;
-  static DEFAULTS = {
-    DIARIZE: false,
-    SILDUR: "1.3",
-    SILBUF: 0.2,
-    SILTHR: "-35dB",
-    MODEL_KEYS: {
-      tinyd: "ggml-small.en-tdrz.bin",
-      small: "ggml-small.bin",
-      medium: "ggml-medium.bin"
-    },
-    MODELS: {
-      tinyd: "https://huggingface.co/akashmjn/tinydiarize-whisper.cpp/resolve/main/ggml-small.en-tdrz.bin",
-      small: "https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-small.bin",
-      medium: "https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-medium.bin"
-    }
-  };
+var _WhisperProvider = class _WhisperProvider {
   constructor(config) {
+    __publicField(this, "name", "whisper");
+    __publicField(this, "cacheDir");
     this.cacheDir = config?.cacheDir || `${process.env.HOME}/.cache/whisper-models`;
   }
   async transcribe(params) {
@@ -187,36 +175,44 @@ var WhisperProvider = class _WhisperProvider {
     return modelPath;
   }
 };
+__publicField(_WhisperProvider, "DEFAULTS", {
+  DIARIZE: false,
+  SILDUR: "1.3",
+  SILBUF: 0.2,
+  SILTHR: "-35dB",
+  MODEL_KEYS: {
+    tinyd: "ggml-small.en-tdrz.bin",
+    small: "ggml-small.bin",
+    medium: "ggml-medium.bin"
+  },
+  MODELS: {
+    tinyd: "https://huggingface.co/akashmjn/tinydiarize-whisper.cpp/resolve/main/ggml-small.en-tdrz.bin",
+    small: "https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-small.bin",
+    medium: "https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-medium.bin"
+  }
+});
+var WhisperProvider = _WhisperProvider;
 var VOXTRAL_LIMITS = {
   /** Maximum audio duration in seconds (3 hours for Voxtral Transcribe 2) */
   maxAudioDurationSec: 3 * 60 * 60,
   // 10800 seconds = 3 hours
-  /** Recommended max duration before splitting (for reliability) */
-  recommendedMaxDurationSec: 30 * 60,
-  // 30 minutes
   /** Maximum context biasing words/phrases */
   maxContextBiasingTerms: 100,
   /** Maximum file size in bytes (1GB) */
   maxFileSizeBytes: 1024 * 1024 * 1024
 };
 var MistralProvider = class {
-  name = "mistral";
-  maxAudioDurationSec = VOXTRAL_LIMITS.maxAudioDurationSec;
-  /**
-   * Check if audio duration exceeds recommended limits
-   */
-  static shouldSplit(durationSec) {
-    return durationSec > VOXTRAL_LIMITS.recommendedMaxDurationSec;
-  }
-  /**
-   * Get the recommended max segment duration for splitting
-   */
-  static getRecommendedMaxSegment() {
-    return VOXTRAL_LIMITS.recommendedMaxDurationSec;
+  constructor() {
+    __publicField(this, "name", "mistral");
+    __publicField(this, "maxAudioDurationSec", VOXTRAL_LIMITS.maxAudioDurationSec);
   }
   async transcribe(params) {
     if (params.language && params.timestampGranularity) {
-      return { text: "", error: "Cannot use both language and timestampGranularity (Mistral API limitation)" };
+      throw new Error("Cannot use both language and timestampGranularity (Mistral API limitation)");
+    }
+    const diarize = params.diarize ?? true;
+    if (diarize && params.timestampGranularity === "word") {
+      throw new Error('When diarize is set to true, the timestamp granularity must be set to ["segment"], got ["word"]');
     }
     const formData = new FormData();
     if (params.audioUrl) {
@@ -242,7 +238,6 @@ var MistralProvider = class {
     if (params.language) {
       formData.append("language", params.language);
     }
-    const diarize = params.diarize ?? true;
     if (diarize) {
       formData.append("diarize", "true");
     }
@@ -278,7 +273,9 @@ var MistralProvider = class {
   }
 };
 var GreenPTProvider = class {
-  name = "greenpt";
+  constructor() {
+    __publicField(this, "name", "greenpt");
+  }
   async transcribe(params) {
     if (!params.apiKey) {
       return { text: "", error: "API key is required for GreenPT provider" };
@@ -866,7 +863,7 @@ function createMistralTranscriber(config) {
         language,
         model = defaultModel,
         diarize = true,
-        timestamps = language ? void 0 : "word",
+        timestamps = language ? void 0 : "segment",
         autoSplit,
         splitOutputDir,
         logger: customLogger,
@@ -874,7 +871,7 @@ function createMistralTranscriber(config) {
       } = options;
       const log = customLogger || defaultLogger;
       if (verbose) log.debug = log.info;
-      const maxDuration = VOXTRAL_LIMITS.recommendedMaxDurationSec;
+      const maxDuration = VOXTRAL_LIMITS.maxAudioDurationSec;
       if (audioBuffer) {
         log.info(`Transcribing from buffer (${(audioBuffer.length / 1024 / 1024).toFixed(2)} MB)`);
         const result = await provider.transcribe({
@@ -1003,6 +1000,7 @@ export {
   autoSplitAudio,
   createMistralTranscriber,
   createProvider,
+  detectAudioMimeType,
   detectSilenceRegions,
   findOptimalSplitPoints,
   formatMergedText,

package/dist/providers.min.js

@@ -1,3 +1,7 @@
+var __defProp = Object.defineProperty;
+var __defNormalProp = (obj, key, value) => key in obj ? __defProp(obj, key, { enumerable: true, configurable: true, writable: true, value }) : obj[key] = value;
+var __publicField = (obj, key, value) => __defNormalProp(obj, typeof key !== "symbol" ? key + "" : key, value);
+
 // src/utils/transcription/providers.ts
 import { spawn } from "child_process";
 import * as fs from "fs";
@@ -57,26 +61,10 @@ function createProvider(providerName, config) {
       throw new Error(`Unknown provider: ${providerName}`);
   }
 }
-var WhisperProvider = class _WhisperProvider {
-  name = "whisper";
-  cacheDir;
-  static DEFAULTS = {
-    DIARIZE: false,
-    SILDUR: "1.3",
-    SILBUF: 0.2,
-    SILTHR: "-35dB",
-    MODEL_KEYS: {
-      tinyd: "ggml-small.en-tdrz.bin",
-      small: "ggml-small.bin",
-      medium: "ggml-medium.bin"
-    },
-    MODELS: {
-      tinyd: "https://huggingface.co/akashmjn/tinydiarize-whisper.cpp/resolve/main/ggml-small.en-tdrz.bin",
-      small: "https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-small.bin",
-      medium: "https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-medium.bin"
-    }
-  };
+var _WhisperProvider = class _WhisperProvider {
   constructor(config) {
+    __publicField(this, "name", "whisper");
+    __publicField(this, "cacheDir");
     this.cacheDir = config?.cacheDir || `${process.env.HOME}/.cache/whisper-models`;
   }
   async transcribe(params) {
@@ -187,36 +175,44 @@ var WhisperProvider = class _WhisperProvider {
     return modelPath;
   }
 };
+__publicField(_WhisperProvider, "DEFAULTS", {
+  DIARIZE: false,
+  SILDUR: "1.3",
+  SILBUF: 0.2,
+  SILTHR: "-35dB",
+  MODEL_KEYS: {
+    tinyd: "ggml-small.en-tdrz.bin",
+    small: "ggml-small.bin",
+    medium: "ggml-medium.bin"
+  },
+  MODELS: {
+    tinyd: "https://huggingface.co/akashmjn/tinydiarize-whisper.cpp/resolve/main/ggml-small.en-tdrz.bin",
+    small: "https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-small.bin",
+    medium: "https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-medium.bin"
+  }
+});
+var WhisperProvider = _WhisperProvider;
 var VOXTRAL_LIMITS = {
   /** Maximum audio duration in seconds (3 hours for Voxtral Transcribe 2) */
   maxAudioDurationSec: 3 * 60 * 60,
   // 10800 seconds = 3 hours
-  /** Recommended max duration before splitting (for reliability) */
-  recommendedMaxDurationSec: 30 * 60,
-  // 30 minutes
   /** Maximum context biasing words/phrases */
   maxContextBiasingTerms: 100,
   /** Maximum file size in bytes (1GB) */
   maxFileSizeBytes: 1024 * 1024 * 1024
 };
 var MistralProvider = class {
-  name = "mistral";
-  maxAudioDurationSec = VOXTRAL_LIMITS.maxAudioDurationSec;
-  /**
-   * Check if audio duration exceeds recommended limits
-   */
-  static shouldSplit(durationSec) {
-    return durationSec > VOXTRAL_LIMITS.recommendedMaxDurationSec;
-  }
-  /**
-   * Get the recommended max segment duration for splitting
-   */
-  static getRecommendedMaxSegment() {
-    return VOXTRAL_LIMITS.recommendedMaxDurationSec;
+  constructor() {
+    __publicField(this, "name", "mistral");
+    __publicField(this, "maxAudioDurationSec", VOXTRAL_LIMITS.maxAudioDurationSec);
   }
   async transcribe(params) {
     if (params.language && params.timestampGranularity) {
-      return { text: "", error: "Cannot use both language and timestampGranularity (Mistral API limitation)" };
+      throw new Error("Cannot use both language and timestampGranularity (Mistral API limitation)");
+    }
+    const diarize = params.diarize ?? true;
+    if (diarize && params.timestampGranularity === "word") {
+      throw new Error('When diarize is set to true, the timestamp granularity must be set to ["segment"], got ["word"]');
     }
     const formData = new FormData();
     if (params.audioUrl) {
@@ -242,7 +238,6 @@ var MistralProvider = class {
     if (params.language) {
       formData.append("language", params.language);
     }
-    const diarize = params.diarize ?? true;
     if (diarize) {
       formData.append("diarize", "true");
     }
@@ -278,7 +273,9 @@ var MistralProvider = class {
   }
 };
 var GreenPTProvider = class {
-  name = "greenpt";
+  constructor() {
+    __publicField(this, "name", "greenpt");
+  }
   async transcribe(params) {
     if (!params.apiKey) {
       return { text: "", error: "API key is required for GreenPT provider" };
@@ -788,7 +785,7 @@ function createMistralTranscriber(config) {
         language,
         model = defaultModel,
         diarize = true,
-        timestamps = language ? void 0 : "word",
+        timestamps = language ? void 0 : "segment",
         autoSplit,
         splitOutputDir,
         logger: customLogger,
@@ -796,7 +793,7 @@ function createMistralTranscriber(config) {
       } = options;
       const log = customLogger || defaultLogger;
       if (verbose) log.debug = log.info;
-      const maxDuration = VOXTRAL_LIMITS.recommendedMaxDurationSec;
+      const maxDuration = VOXTRAL_LIMITS.maxAudioDurationSec;
       if (audioBuffer) {
         log.info(`Transcribing from buffer (${(audioBuffer.length / 1024 / 1024).toFixed(2)} MB)`);
         const result = await provider.transcribe({

package/dist/realtime.d.ts

@@ -4,7 +4,10 @@
  * This module provides a simple, event-driven interface for streaming audio
  * transcription using Mistral's realtime WebSocket API.
  *
- * @example
+ * **Node.js only** - Browser support is currently disabled due to WebSocket
+ * authentication limitations with Mistral API.
+ *
+ * @example Node.js
  * ```typescript
  * import {
  *   createRealtimeTranscriber,
@@ -15,7 +18,7 @@
  *   apiKey: process.env.MISTRAL_API_KEY,
  * })
  *
- * const { stream, stop } = captureAudioFromMicrophone(16000)
+ * const { stream, stop } = await captureAudioFromMicrophone(16000)
  *
  * try {
  *   for await (const event of transcriber.transcribe(stream)) {
@@ -33,6 +36,5 @@
  *
  * @module @wovin/tranz/realtime
  */
-export { createRealtimeTranscriber, captureAudioFromMicrophone, type RealtimeEvent, type RealtimeConfig, type RealtimeTranscriber, type TranscribeOptions, type AudioFormat, type AudioCaptureResult, type SessionCreatedEvent, type SessionUpdatedEvent, type TranscriptionTextDeltaEvent, type TranscriptionLanguageEvent, type TranscriptionSegmentEvent, type TranscriptionDoneEvent, type ErrorEvent, } from "./utils/transcription/realtime.js";
-export { AudioEncoding } from "@mistralai/mistralai/extra/realtime";
+export { createRealtimeTranscriber, captureAudioFromMicrophone, captureAudioFromBrowser, AudioEncoding, type RealtimeEvent, type RealtimeConfig, type RealtimeTranscriber, type TranscribeOptions, type AudioFormat, type AudioCaptureResult, type SessionCreatedEvent, type SessionUpdatedEvent, type TranscriptionTextDeltaEvent, type TranscriptionLanguageEvent, type TranscriptionSegmentEvent, type TranscriptionDoneEvent, type ErrorEvent, } from "./utils/transcription/realtime.js";
 //# sourceMappingURL=realtime.d.ts.map

package/dist/realtime.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"realtime.d.ts","sourceRoot":"","sources":["../src/realtime.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAkCG;AAEH,OAAO,EACL,yBAAyB,EACzB,0BAA0B,EAC1B,KAAK,aAAa,EAClB,KAAK,cAAc,EACnB,KAAK,mBAAmB,EACxB,KAAK,iBAAiB,EACtB,KAAK,WAAW,EAChB,KAAK,kBAAkB,EACvB,KAAK,mBAAmB,EACxB,KAAK,mBAAmB,EACxB,KAAK,2BAA2B,EAChC,KAAK,0BAA0B,EAC/B,KAAK,yBAAyB,EAC9B,KAAK,sBAAsB,EAC3B,KAAK,UAAU,GAChB,MAAM,mCAAmC,CAAC;AAG3C,OAAO,EAAE,aAAa,EAAE,MAAM,qCAAqC,CAAC"}
+{"version":3,"file":"realtime.d.ts","sourceRoot":"","sources":["../src/realtime.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAqCG;AAEH,OAAO,EACL,yBAAyB,EACzB,0BAA0B,EAC1B,uBAAuB,EACvB,aAAa,EACb,KAAK,aAAa,EAClB,KAAK,cAAc,EACnB,KAAK,mBAAmB,EACxB,KAAK,iBAAiB,EACtB,KAAK,WAAW,EAChB,KAAK,kBAAkB,EACvB,KAAK,mBAAmB,EACxB,KAAK,mBAAmB,EACxB,KAAK,2BAA2B,EAChC,KAAK,0BAA0B,EAC/B,KAAK,yBAAyB,EAC9B,KAAK,sBAAsB,EAC3B,KAAK,UAAU,GAChB,MAAM,mCAAmC,CAAC"}

package/dist/realtime.min.js

@@ -1,32 +1,309 @@
+// src/utils/transcription/runtime.ts
+async function getWebSocketImpl() {
+  const isBrowser = typeof globalThis !== "undefined" && (typeof globalThis.document !== "undefined" || typeof globalThis.navigator !== "undefined");
+  if (isBrowser && typeof globalThis.WebSocket !== "undefined") {
+    return globalThis.WebSocket;
+  }
+  if (!isBrowser) {
+    try {
+      const WS = await import("ws");
+      return WS.default || WS;
+    } catch (err) {
+      throw new Error(
+        "WebSocket not available. In Node.js, install 'ws' package: npm install ws"
+      );
+    }
+  }
+  throw new Error(
+    "WebSocket not available in this environment"
+  );
+}
+
 // src/utils/transcription/realtime.ts
-import { spawn } from "child_process";
-import {
-  AudioEncoding,
-  RealtimeTranscription
-} from "@mistralai/mistralai/extra/realtime";
+var AudioEncoding = /* @__PURE__ */ ((AudioEncoding2) => {
+  AudioEncoding2["PcmS16le"] = "pcm_s16le";
+  AudioEncoding2["PcmS16be"] = "pcm_s16be";
+  AudioEncoding2["PcmU16le"] = "pcm_u16le";
+  AudioEncoding2["PcmU16be"] = "pcm_u16be";
+  AudioEncoding2["PcmS24le"] = "pcm_s24le";
+  AudioEncoding2["PcmS24be"] = "pcm_s24be";
+  AudioEncoding2["PcmU24le"] = "pcm_u24le";
+  AudioEncoding2["PcmU24be"] = "pcm_u24be";
+  AudioEncoding2["PcmS32le"] = "pcm_s32le";
+  AudioEncoding2["PcmS32be"] = "pcm_s32be";
+  AudioEncoding2["PcmU32le"] = "pcm_u32le";
+  AudioEncoding2["PcmU32be"] = "pcm_u32be";
+  AudioEncoding2["PcmF32le"] = "pcm_f32le";
+  AudioEncoding2["PcmF32be"] = "pcm_f32be";
+  AudioEncoding2["PcmF64le"] = "pcm_f64le";
+  AudioEncoding2["PcmF64be"] = "pcm_f64be";
+  return AudioEncoding2;
+})(AudioEncoding || {});
 function createRealtimeTranscriber(config) {
+  const isBrowser = typeof window !== "undefined" && typeof document !== "undefined" && typeof navigator !== "undefined";
+  if (isBrowser) {
+    throw new Error(
+      "Realtime transcription is not yet supported in browsers. Browser WebSocket API does not support authentication headers required by Mistral API. Use this API in Node.js or server-side environments only. See: https://github.com/wovin/tranz/issues"
+    );
+  }
   const model = config.model ?? "voxtral-mini-transcribe-realtime-2602";
   const baseUrl = config.baseUrl ?? "wss://api.mistral.ai";
-  const client = new RealtimeTranscription({
-    apiKey: config.apiKey,
-    serverURL: baseUrl
-  });
   return {
     async *transcribe(audioStream, options) {
       const audioFormat = {
-        encoding: options?.audioFormat?.encoding ?? AudioEncoding.PcmS16le,
+        encoding: options?.audioFormat?.encoding ?? "pcm_s16le" /* PcmS16le */,
         sampleRate: options?.audioFormat?.sampleRate ?? 16e3
       };
-      const eventStream = client.transcribeStream(audioStream, model, {
+      const connection = await createConnection(
+        config.apiKey,
+        baseUrl,
+        model,
         audioFormat
-      });
-      for await (const event of eventStream) {
-        yield event;
+      );
+      try {
+        let stopRequested = false;
+        const sendAudioTask = (async () => {
+          try {
+            for await (const chunk of audioStream) {
+              if (stopRequested || connection.isClosed) {
+                break;
+              }
+              await connection.sendAudio(chunk);
+            }
+          } finally {
+            await connection.endAudio();
+          }
+        })();
+        for await (const event of connection.events()) {
+          yield event;
+          if (event.type === "transcription.done" || event.type === "error") {
+            break;
+          }
+        }
+        await sendAudioTask;
+      } finally {
+        await connection.close();
+        const maybeReturn = audioStream.return;
+        if (typeof maybeReturn === "function") {
+          await maybeReturn.call(audioStream);
+        }
       }
     }
   };
 }
-function captureAudioFromMicrophone(sampleRate = 16e3) {
+async function createConnection(apiKey, baseUrl, model, audioFormat) {
+  const WebSocketImpl = await getWebSocketImpl();
+  const wsUrl = buildWebSocketUrl(baseUrl, model, apiKey);
+  const isNodeWs = typeof process !== "undefined" && process.versions?.node;
+  const ws = isNodeWs ? new WebSocketImpl(wsUrl, {
+    headers: {
+      Authorization: `Bearer ${apiKey}`
+    }
+  }) : new WebSocketImpl(wsUrl);
+  const session = await waitForSession(ws);
+  let closed = false;
+  const websocket = ws;
+  const connection = {
+    get isClosed() {
+      return closed || websocket.readyState === 2 || websocket.readyState === 3;
+    },
+    async *events() {
+      const queue = [];
+      let resolver = null;
+      let done = false;
+      const push = (item) => {
+        if (done) return;
+        if (resolver) {
+          const resolve = resolver;
+          resolver = null;
+          resolve(item);
+          return;
+        }
+        queue.push(item);
+      };
+      const handleMessage = (event) => {
+        push({ kind: "message", data: event.data });
+      };
+      const handleClose = () => {
+        closed = true;
+        push({ kind: "close" });
+      };
+      const handleError = (event) => {
+        push({
+          kind: "error",
+          error: new Error("WebSocket connection error")
+        });
+      };
+      websocket.addEventListener("message", handleMessage);
+      websocket.addEventListener("close", handleClose);
+      websocket.addEventListener("error", handleError);
+      try {
+        while (true) {
+          const item = queue.length > 0 ? queue.shift() : await new Promise((resolve) => {
+            resolver = resolve;
+          });
+          if (item.kind === "close") break;
+          if (item.kind === "error") {
+            const error = item.error ?? new Error("WebSocket connection error");
+            yield {
+              type: "error",
+              error: { message: error.message }
+            };
+            continue;
+          }
+          const event = parseRealtimeEvent(item.data);
+          yield event;
+        }
+      } finally {
+        done = true;
+        websocket.removeEventListener("message", handleMessage);
+        websocket.removeEventListener("close", handleClose);
+        websocket.removeEventListener("error", handleError);
+        if (resolver !== null) {
+          const resolve = resolver;
+          resolver = null;
+          resolve({ kind: "close" });
+        }
+      }
+    },
+    async sendAudio(chunk) {
+      if (connection.isClosed) {
+        throw new Error("Connection is closed");
+      }
+      const base64Audio = arrayBufferToBase64(chunk);
+      const message = {
+        type: "input_audio.append",
+        audio: base64Audio
+      };
+      await sendJson(websocket, message);
+    },
+    async endAudio() {
+      if (connection.isClosed) return;
+      await sendJson(websocket, { type: "input_audio.end" });
+    },
+    async close() {
+      if (closed) return;
+      closed = true;
+      if (websocket.readyState === 3) return;
+      await new Promise((resolve) => {
+        const finalize = () => {
+          websocket.removeEventListener("close", finalize);
+          resolve();
+        };
+        websocket.addEventListener("close", finalize);
+        websocket.close(1e3, "");
+      });
+    }
+  };
+  return connection;
+}
+function buildWebSocketUrl(baseUrl, model, apiKey) {
+  const url = new URL("v1/audio/transcriptions/realtime", baseUrl);
+  url.searchParams.set("model", model);
+  return url.toString();
+}
+async function waitForSession(ws) {
+  return new Promise((resolve, reject) => {
+    const timeout = setTimeout(() => {
+      cleanup();
+      ws.close();
+      reject(new Error("Timeout waiting for session creation"));
+    }, 1e4);
+    const cleanup = () => {
+      clearTimeout(timeout);
+      ws.removeEventListener("message", handleMessage);
+      ws.removeEventListener("close", handleClose);
+      ws.removeEventListener("error", handleError);
+    };
+    const handleMessage = (event) => {
+      try {
+        const parsed = parseRealtimeEvent(event.data);
+        if (parsed.type === "session.created") {
+          cleanup();
+          resolve(parsed);
+        } else if (parsed.type === "error") {
+          cleanup();
+          ws.close();
+          reject(
+            new Error(
+              `Realtime transcription error: ${JSON.stringify(parsed.error)}`
+            )
+          );
+        }
+      } catch (err) {
+        cleanup();
+        ws.close();
+        reject(err);
+      }
+    };
+    const handleClose = () => {
+      cleanup();
+      reject(new Error("WebSocket closed during handshake"));
+    };
+    const handleError = () => {
+      cleanup();
+      reject(new Error("WebSocket error during handshake"));
+    };
+    ws.addEventListener("message", handleMessage);
+    ws.addEventListener("close", handleClose);
+    ws.addEventListener("error", handleError);
+  });
+}
+function parseRealtimeEvent(data) {
+  try {
+    const text = typeof data === "string" ? data : new TextDecoder().decode(data);
+    const payload = JSON.parse(text);
+    if (typeof payload.type !== "string") {
+      return {
+        type: "error",
+        error: { message: "Invalid event: missing type" }
+      };
+    }
+    return payload;
+  } catch (err) {
+    return {
+      type: "error",
+      error: { message: `Failed to parse event: ${err}` }
+    };
+  }
+}
+async function sendJson(ws, payload) {
+  return new Promise((resolve, reject) => {
+    const message = JSON.stringify(payload);
+    if (typeof ws.send === "function") {
+      const send = ws.send.bind(ws);
+      try {
+        send(message, (err) => {
+          if (err) reject(err);
+          else resolve();
+        });
+      } catch {
+        ws.send(message);
+        resolve();
+      }
+    } else {
+      ws.send(message);
+      resolve();
+    }
+  });
+}
+function arrayBufferToBase64(buffer) {
+  if (typeof btoa !== "undefined") {
+    const binary = Array.from(buffer).map((byte) => String.fromCharCode(byte)).join("");
+    return btoa(binary);
+  }
+  if (typeof Buffer !== "undefined") {
+    return Buffer.from(buffer).toString("base64");
+  }
+  throw new Error("No base64 encoding available");
+}
+async function captureAudioFromMicrophone(sampleRate = 16e3) {
+  if (typeof process === "undefined" || !process.versions?.node) {
+    throw new Error(
+      "captureAudioFromMicrophone() is Node.js only. Use captureAudioFromBrowser() in browsers."
+    );
+  }
+  const { spawn } = await import("child_process");
   const recorder = spawn(
     "rec",
     [
@@ -53,8 +330,7 @@ function captureAudioFromMicrophone(sampleRate = 16e3) {
     { stdio: ["ignore", "pipe", "ignore"] }
   );
   recorder.on("error", (err) => {
-    const error = err;
-    if (error.code === "ENOENT") {
+    if (err.code === "ENOENT") {
       console.error(
         "\nError: 'rec' command not found. Please install SoX:",
         "\n  macOS: brew install sox",
@@ -85,11 +361,88 @@ function captureAudioFromMicrophone(sampleRate = 16e3) {
   };
   return { stream, stop };
 }
-
-// src/realtime.ts
-import { AudioEncoding as AudioEncoding2 } from "@mistralai/mistralai/extra/realtime";
+async function captureAudioFromBrowser(sampleRate = 16e3) {
+  throw new Error(
+    "Browser realtime transcription is not yet supported. Browser WebSocket API does not support authentication headers required by Mistral API. Use captureAudioFromMicrophone() in Node.js environments instead."
+  );
+  if (typeof navigator === "undefined" || !navigator.mediaDevices) {
+    throw new Error(
+      "captureAudioFromBrowser() requires a browser environment with getUserMedia support"
+    );
+  }
+  const mediaStream = await navigator.mediaDevices.getUserMedia({
+    audio: {
+      channelCount: 1,
+      sampleRate,
+      echoCancellation: true,
+      noiseSuppression: true
+    }
+  });
+  const audioContext = new AudioContext({ sampleRate });
+  const source = audioContext.createMediaStreamSource(mediaStream);
+  const processor = audioContext.createScriptProcessor(4096, 1, 1);
+  let stopped = false;
+  const chunks = [];
+  let resolver = null;
+  processor.onaudioprocess = (event) => {
+    if (stopped) return;
+    const inputData = event.inputBuffer.getChannelData(0);
+    const pcm16 = new Int16Array(inputData.length);
+    for (let i = 0; i < inputData.length; i++) {
+      const sample = Math.max(-1, Math.min(1, inputData[i]));
+      pcm16[i] = sample < 0 ? sample * 32768 : sample * 32767;
+    }
+    const uint8 = new Uint8Array(pcm16.length * 2);
+    for (let i = 0; i < pcm16.length; i++) {
+      uint8[i * 2] = pcm16[i] & 255;
+      uint8[i * 2 + 1] = pcm16[i] >> 8 & 255;
+    }
+    if (resolver) {
+      const resolve = resolver;
+      resolver = null;
+      resolve({ value: uint8, done: false });
+    } else {
+      chunks.push(pcm16);
+    }
+  };
+  source.connect(processor);
+  processor.connect(audioContext.destination);
+  const stream = (async function* () {
+    try {
+      while (!stopped) {
+        if (chunks.length > 0) {
+          const pcm16 = chunks.shift();
+          const uint8 = new Uint8Array(pcm16.length * 2);
+          for (let i = 0; i < pcm16.length; i++) {
+            uint8[i * 2] = pcm16[i] & 255;
+            uint8[i * 2 + 1] = pcm16[i] >> 8 & 255;
+          }
+          yield uint8;
+        } else {
+          await new Promise((resolve) => {
+            resolver = resolve;
+          });
+        }
+      }
+    } finally {
+      processor.disconnect();
+      source.disconnect();
+      mediaStream.getTracks().forEach((track) => track.stop());
+      await audioContext.close();
+    }
+  })();
+  const stop = () => {
+    stopped = true;
+    if (resolver) {
+      resolver({ value: void 0, done: true });
+      resolver = null;
+    }
+  };
+  return { stream, stop };
+}
 export {
-  AudioEncoding2 as AudioEncoding,
+  AudioEncoding,
+  captureAudioFromBrowser,
   captureAudioFromMicrophone,
   createRealtimeTranscriber
 };

package/dist/utils/transcription/providers.d.ts

@@ -115,8 +115,6 @@ export declare class WhisperProvider implements TranscriptionProvider {
 export declare const VOXTRAL_LIMITS: {
     /** Maximum audio duration in seconds (3 hours for Voxtral Transcribe 2) */
     maxAudioDurationSec: number;
-    /** Recommended max duration before splitting (for reliability) */
-    recommendedMaxDurationSec: number;
     /** Maximum context biasing words/phrases */
     maxContextBiasingTerms: number;
     /** Maximum file size in bytes (1GB) */
@@ -125,14 +123,6 @@ export declare const VOXTRAL_LIMITS: {
 export declare class MistralProvider implements TranscriptionProvider {
     name: string;
     maxAudioDurationSec: number;
-    /**
-     * Check if audio duration exceeds recommended limits
-     */
-    static shouldSplit(durationSec: number): boolean;
-    /**
-     * Get the recommended max segment duration for splitting
-     */
-    static getRecommendedMaxSegment(): number;
     transcribe(params: TranscribeParams): Promise<TranscriptionResult>;
 }
 /**

package/dist/utils/transcription/providers.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"providers.d.ts","sourceRoot":"","sources":["../../../src/utils/transcription/providers.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAUH;;;GAGG;AACH,MAAM,WAAW,mBAAmB;IAClC,mCAAmC;IACnC,IAAI,EAAE,MAAM,CAAA;IACZ,sEAAsE;IACtE,WAAW,CAAC,EAAE,GAAG,CAAA;IACjB,4CAA4C;IAC5C,KAAK,CAAC,EAAE,MAAM,CAAA;IACd,kDAAkD;IAClD,UAAU,CAAC,EAAE,MAAM,CAAA;IACnB,iDAAiD;IACjD,KAAK,CAAC,EAAE,GAAG,EAAE,CAAA;IACb,mCAAmC;IACnC,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,0CAA0C;IAC1C,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,6DAA6D;IAC7D,KAAK,CAAC,EAAE,MAAM,CAAA;CACf;AAED;;;GAGG;AACH,MAAM,WAAW,qBAAqB;IACpC,+BAA+B;IAC/B,IAAI,EAAE,MAAM,CAAA;IACZ,+DAA+D;IAC/D,mBAAmB,CAAC,EAAE,MAAM,CAAA;IAC5B;;;;OAIG;IACH,UAAU,CAAC,MAAM,EAAE,gBAAgB,GAAG,OAAO,CAAC,mBAAmB,CAAC,CAAA;CACnE;AAED;;;GAGG;AACH,MAAM,WAAW,gBAAgB;IAC/B,2CAA2C;IAC3C,SAAS,CAAC,EAAE,MAAM,CAAA;IAClB,iCAAiC;IACjC,WAAW,CAAC,EAAE,MAAM,CAAA;IACpB,gEAAgE;IAChE,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,iDAAiD;IACjD,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,yDAAyD;IACzD,KAAK,CAAC,EAAE,MAAM,CAAA;IACd,yDAAyD;IACzD,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,qDAAqD;IACrD,MAAM,CAAC,EAAE,MAAM,CAAA;IACf,oDAAoD;IACpD,OAAO,CAAC,EAAE,OAAO,CAAA;IACjB,iEAAiE;IACjE,oBAAoB,CAAC,EAAE,SAAS,GAAG,MAAM,CAAA;IACzC,4CAA4C;IAC5C,SAAS,CAAC,EAAE,MAAM,CAAA;IAClB,sDAAsD;IACtD,SAAS,CAAC,EAAE,MAAM,CAAA;IAClB,oCAAoC;IACpC,MAAM,CAAC,EAAE,GAAG,CAAA;CACb;AAED;;GAEG;AACH,MAAM,MAAM,YAAY,GAAG,SAAS,GAAG,SAAS,GAAG,SAAS,CAAA;AAE5D;;;;;GAKG;AACH,wBAAgB,cAAc,CAAC,YAAY,EAAE,YAAY,EAAE,MAAM,CAAC,EAAE,GAAG,GAAG,qBAAqB,CAW9F;AASD;;;GAGG;AACH,qBAAa,eAAgB,YAAW,qBAAqB;IAC3D,IAAI,SAAY;IAEhB,OAAO,CAAC,QAAQ,CAAQ;IAExB,MAAM,CAAC,QAAQ;;;;;;;;;;;;;;;MAkBd;gBAEW,MAAM,CAAC,EAAE,GAAG;IAKlB,UAAU,CAAC,MAAM,EAAE,gBAAgB,GAAG,OAAO,CAAC,mBAAmB,CAAC;YA+F1D,4BAA4B;CAkC3C;AAED;;;GAGG;AACH,eAAO,MAAM,cAAc;IACzB,2EAA2E;;IAE3E,kEAAkE;;IAElE,4CAA4C;;IAE5C,uCAAuC;;CAExC,CAAA;AAED,qBAAa,eAAgB,YAAW,qBAAqB;IAC3D,IAAI,SAAY;IAChB,mBAAmB,SAAqC;IAExD;;OAEG;IACH,MAAM,CAAC,WAAW,CAAC,WAAW,EAAE,MAAM,GAAG,OAAO;IAIhD;;OAEG;IACH,MAAM,CAAC,wBAAwB,IAAI,MAAM;IAInC,UAAU,CAAC,MAAM,EAAE,gBAAgB,GAAG,OAAO,CAAC,mBAAmB,CAAC;CAmGzE;AAED;;;GAGG;AACH,qBAAa,eAAgB,YAAW,qBAAqB;IAC3D,IAAI,SAAY;IAEV,UAAU,CAAC,MAAM,EAAE,gBAAgB,GAAG,OAAO,CAAC,mBAAmB,CAAC;CAkGzE"}
+{"version":3,"file":"providers.d.ts","sourceRoot":"","sources":["../../../src/utils/transcription/providers.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAUH;;;GAGG;AACH,MAAM,WAAW,mBAAmB;IAClC,mCAAmC;IACnC,IAAI,EAAE,MAAM,CAAA;IACZ,sEAAsE;IACtE,WAAW,CAAC,EAAE,GAAG,CAAA;IACjB,4CAA4C;IAC5C,KAAK,CAAC,EAAE,MAAM,CAAA;IACd,kDAAkD;IAClD,UAAU,CAAC,EAAE,MAAM,CAAA;IACnB,iDAAiD;IACjD,KAAK,CAAC,EAAE,GAAG,EAAE,CAAA;IACb,mCAAmC;IACnC,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,0CAA0C;IAC1C,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,6DAA6D;IAC7D,KAAK,CAAC,EAAE,MAAM,CAAA;CACf;AAED;;;GAGG;AACH,MAAM,WAAW,qBAAqB;IACpC,+BAA+B;IAC/B,IAAI,EAAE,MAAM,CAAA;IACZ,+DAA+D;IAC/D,mBAAmB,CAAC,EAAE,MAAM,CAAA;IAC5B;;;;OAIG;IACH,UAAU,CAAC,MAAM,EAAE,gBAAgB,GAAG,OAAO,CAAC,mBAAmB,CAAC,CAAA;CACnE;AAED;;;GAGG;AACH,MAAM,WAAW,gBAAgB;IAC/B,2CAA2C;IAC3C,SAAS,CAAC,EAAE,MAAM,CAAA;IAClB,iCAAiC;IACjC,WAAW,CAAC,EAAE,MAAM,CAAA;IACpB,gEAAgE;IAChE,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,iDAAiD;IACjD,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,yDAAyD;IACzD,KAAK,CAAC,EAAE,MAAM,CAAA;IACd,yDAAyD;IACzD,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,qDAAqD;IACrD,MAAM,CAAC,EAAE,MAAM,CAAA;IACf,oDAAoD;IACpD,OAAO,CAAC,EAAE,OAAO,CAAA;IACjB,iEAAiE;IACjE,oBAAoB,CAAC,EAAE,SAAS,GAAG,MAAM,CAAA;IACzC,4CAA4C;IAC5C,SAAS,CAAC,EAAE,MAAM,CAAA;IAClB,sDAAsD;IACtD,SAAS,CAAC,EAAE,MAAM,CAAA;IAClB,oCAAoC;IACpC,MAAM,CAAC,EAAE,GAAG,CAAA;CACb;AAED;;GAEG;AACH,MAAM,MAAM,YAAY,GAAG,SAAS,GAAG,SAAS,GAAG,SAAS,CAAA;AAE5D;;;;;GAKG;AACH,wBAAgB,cAAc,CAAC,YAAY,EAAE,YAAY,EAAE,MAAM,CAAC,EAAE,GAAG,GAAG,qBAAqB,CAW9F;AASD;;;GAGG;AACH,qBAAa,eAAgB,YAAW,qBAAqB;IAC3D,IAAI,SAAY;IAEhB,OAAO,CAAC,QAAQ,CAAQ;IAExB,MAAM,CAAC,QAAQ;;;;;;;;;;;;;;;MAkBd;gBAEW,MAAM,CAAC,EAAE,GAAG;IAKlB,UAAU,CAAC,MAAM,EAAE,gBAAgB,GAAG,OAAO,CAAC,mBAAmB,CAAC;YA+F1D,4BAA4B;CAkC3C;AAED;;;GAGG;AACH,eAAO,MAAM,cAAc;IACzB,2EAA2E;;IAE3E,4CAA4C;;IAE5C,uCAAuC;;CAExC,CAAA;AAED,qBAAa,eAAgB,YAAW,qBAAqB;IAC3D,IAAI,SAAY;IAChB,mBAAmB,SAAqC;IAElD,UAAU,CAAC,MAAM,EAAE,gBAAgB,GAAG,OAAO,CAAC,mBAAmB,CAAC;CAwGzE;AAED;;;GAGG;AACH,qBAAa,eAAgB,YAAW,qBAAqB;IAC3D,IAAI,SAAY;IAEV,UAAU,CAAC,MAAM,EAAE,gBAAgB,GAAG,OAAO,CAAC,mBAAmB,CAAC;CAkGzE"}

package/dist/utils/transcription/realtime.d.ts

@@ -4,6 +4,8 @@
  * Provides a simple, event-driven interface for streaming audio transcription.
  * Users provide audio as AsyncIterable<Uint8Array> and receive typed events.
  *
+ * Browser-compatible: Uses native WebSocket in browsers/Deno, 'ws' package in Node.js
+ *
  * @example
  * ```typescript
  * import { createRealtimeTranscriber } from '@wovin/tranz/realtime'
@@ -19,7 +21,27 @@
  * }
  * ```
  */
-import { AudioEncoding } from "@mistralai/mistralai/extra/realtime";
+/**
+ * Audio encoding formats supported by the transcription service
+ */
+export declare enum AudioEncoding {
+    PcmS16le = "pcm_s16le",
+    PcmS16be = "pcm_s16be",
+    PcmU16le = "pcm_u16le",
+    PcmU16be = "pcm_u16be",
+    PcmS24le = "pcm_s24le",
+    PcmS24be = "pcm_s24be",
+    PcmU24le = "pcm_u24le",
+    PcmU24be = "pcm_u24be",
+    PcmS32le = "pcm_s32le",
+    PcmS32be = "pcm_s32be",
+    PcmU32le = "pcm_u32le",
+    PcmU32be = "pcm_u32be",
+    PcmF32le = "pcm_f32le",
+    PcmF32be = "pcm_f32be",
+    PcmF64le = "pcm_f64le",
+    PcmF64be = "pcm_f64be"
+}
 /**
  * Audio format configuration for realtime transcription
  */
@@ -59,6 +81,7 @@ export interface SessionCreatedEvent {
     type: "session.created";
     session: {
         id: string;
+        audioFormat: AudioFormat;
     };
 }
 /**
@@ -112,6 +135,7 @@ export interface ErrorEvent {
     type: "error";
     error: {
         message: string | unknown;
+        code?: string;
     };
 }
 /**
@@ -167,7 +191,7 @@ export interface AudioCaptureResult {
     stop: () => void;
 }
 /**
- * Capture audio from microphone using SoX `rec` command
+ * Capture audio from microphone using SoX `rec` command (Node.js only)
  *
  * Yields PCM 16-bit signed little-endian mono audio chunks suitable for
  * realtime transcription.
@@ -177,6 +201,8 @@ export interface AudioCaptureResult {
  *   - macOS: `brew install sox`
  *   - Linux: `sudo apt install sox`
  *
+ * **Note:** This is Node.js only. For browser audio capture, use `captureAudioFromBrowser()`
+ *
  * @param sampleRate - Sample rate in Hz (default: 16000)
  * @returns Object with audio stream and stop function
  *
@@ -193,5 +219,24 @@ export interface AudioCaptureResult {
  * }
  * ```
  */
-export declare function captureAudioFromMicrophone(sampleRate?: number): AudioCaptureResult;
+export declare function captureAudioFromMicrophone(sampleRate?: number): Promise<AudioCaptureResult>;
+/**
+ * Capture audio from browser microphone using Web Audio API
+ *
+ * **CURRENTLY DISABLED** - Browser support is not available yet due to
+ * WebSocket authentication limitations with Mistral API.
+ *
+ * @deprecated Browser realtime transcription is not yet supported.
+ *             Use captureAudioFromMicrophone() in Node.js instead.
+ *
+ * @param sampleRate - Target sample rate in Hz (default: 16000)
+ * @returns Object with audio stream and stop function
+ *
+ * @throws Error - Always throws as browser mode is disabled
+ *
+ * @todo Enable when Mistral API supports browser WebSocket authentication
+ * @todo Migrate to AudioWorklet for better performance
+ *       See: https://web.dev/patterns/media/microphone-process/
+ */
+export declare function captureAudioFromBrowser(sampleRate?: number): Promise<AudioCaptureResult>;
 //# sourceMappingURL=realtime.d.ts.map

package/dist/utils/transcription/realtime.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"realtime.d.ts","sourceRoot":"","sources":["../../../src/utils/transcription/realtime.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;GAoBG;AAGH,OAAO,EACL,aAAa,EAEd,MAAM,qCAAqC,CAAC;AAM7C;;GAEG;AACH,MAAM,WAAW,WAAW;IAC1B,iDAAiD;IACjD,QAAQ,EAAE,aAAa,CAAC;IACxB,yCAAyC;IACzC,UAAU,EAAE,MAAM,CAAC;CACpB;AAED;;GAEG;AACH,MAAM,WAAW,cAAc;IAC7B,sBAAsB;IACtB,MAAM,EAAE,MAAM,CAAC;IACf,gEAAgE;IAChE,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,yDAAyD;IACzD,OAAO,CAAC,EAAE,MAAM,CAAC;CAClB;AAED;;GAEG;AACH,MAAM,WAAW,iBAAiB;IAChC,2EAA2E;IAC3E,WAAW,CAAC,EAAE,OAAO,CAAC,WAAW,CAAC,CAAC;CACpC;AAED;;;GAGG;AACH,MAAM,MAAM,aAAa,GACrB,mBAAmB,GACnB,mBAAmB,GACnB,2BAA2B,GAC3B,0BAA0B,GAC1B,yBAAyB,GACzB,sBAAsB,GACtB,UAAU,CAAC;AAEf;;GAEG;AACH,MAAM,WAAW,mBAAmB;IAClC,IAAI,EAAE,iBAAiB,CAAC;IACxB,OAAO,EAAE;QACP,EAAE,EAAE,MAAM,CAAC;KACZ,CAAC;CACH;AAED;;GAEG;AACH,MAAM,WAAW,mBAAmB;IAClC,IAAI,EAAE,iBAAiB,CAAC;IACxB,OAAO,EAAE;QACP,WAAW,EAAE,WAAW,CAAC;KAC1B,CAAC;CACH;AAED;;;GAGG;AACH,MAAM,WAAW,2BAA2B;IAC1C,IAAI,EAAE,0BAA0B,CAAC;IACjC,IAAI,EAAE,MAAM,CAAC;CACd;AAED;;GAEG;AACH,MAAM,WAAW,0BAA0B;IACzC,IAAI,EAAE,wBAAwB,CAAC;IAC/B,aAAa,EAAE,MAAM,CAAC;CACvB;AAED;;;GAGG;AACH,MAAM,WAAW,yBAAyB;IACxC,IAAI,EAAE,uBAAuB,CAAC;IAC9B,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,GAAG,CAAC,EAAE,MAAM,CAAC;IACb,IAAI,EAAE,MAAM,CAAC;IACb,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB;AAED;;;GAGG;AACH,MAAM,WAAW,sBAAsB;IACrC,IAAI,EAAE,oBAAoB,CAAC;IAC3B,IAAI,EAAE,MAAM,CAAC;IACb,QAAQ,CAAC,EAAE,MAAM,CAAC;CACnB;AAED;;GAEG;AACH,MAAM,WAAW,UAAU;IACzB,IAAI,EAAE,OAAO,CAAC;IACd,KAAK,EAAE;QACL,OAAO,EAAE,MAAM,GAAG,OAAO,CAAC;KAC3B,CAAC;CACH;AAMD;;GAEG;AACH,MAAM,WAAW,mBAAmB;IAClC;;;;;;;;;;;;;;;;;;;;OAoBG;IACH,UAAU,CACR,WAAW,EAAE,aAAa,CAAC,UAAU,CAAC,EACtC,OAAO,CAAC,EAAE,iBAAiB,GAC1B,aAAa,CAAC,aAAa,CAAC,CAAC;CACjC;AAED;;;;;;;;;;;;;;GAcG;AACH,wBAAgB,yBAAyB,CACvC,MAAM,EAAE,cAAc,GACrB,mBAAmB,CA+BrB;AAMD;;GAEG;AACH,MAAM,WAAW,kBAAkB;IACjC,2CAA2C;IAC3C,MAAM,EAAE,cAAc,CAAC,UAAU,EAAE,IAAI,EAAE,OAAO,CAAC,CAAC;IAClD,qCAAqC;IACrC,IAAI,EAAE,MAAM,IAAI,CAAC;CAClB;AAED;;;;;;;;;;;;;;;;;;;;;;;;;;GA0BG;AACH,wBAAgB,0BAA0B,CACxC,UAAU,GAAE,MAAc,GACzB,kBAAkB,CAuDpB"}
+{"version":3,"file":"realtime.d.ts","sourceRoot":"","sources":["../../../src/utils/transcription/realtime.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;GAsBG;AAQH;;GAEG;AACH,oBAAY,aAAa;IACvB,QAAQ,cAAc;IACtB,QAAQ,cAAc;IACtB,QAAQ,cAAc;IACtB,QAAQ,cAAc;IACtB,QAAQ,cAAc;IACtB,QAAQ,cAAc;IACtB,QAAQ,cAAc;IACtB,QAAQ,cAAc;IACtB,QAAQ,cAAc;IACtB,QAAQ,cAAc;IACtB,QAAQ,cAAc;IACtB,QAAQ,cAAc;IACtB,QAAQ,cAAc;IACtB,QAAQ,cAAc;IACtB,QAAQ,cAAc;IACtB,QAAQ,cAAc;CACvB;AAED;;GAEG;AACH,MAAM,WAAW,WAAW;IAC1B,iDAAiD;IACjD,QAAQ,EAAE,aAAa,CAAC;IACxB,yCAAyC;IACzC,UAAU,EAAE,MAAM,CAAC;CACpB;AAED;;GAEG;AACH,MAAM,WAAW,cAAc;IAC7B,sBAAsB;IACtB,MAAM,EAAE,MAAM,CAAC;IACf,gEAAgE;IAChE,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,yDAAyD;IACzD,OAAO,CAAC,EAAE,MAAM,CAAC;CAClB;AAED;;GAEG;AACH,MAAM,WAAW,iBAAiB;IAChC,2EAA2E;IAC3E,WAAW,CAAC,EAAE,OAAO,CAAC,WAAW,CAAC,CAAC;CACpC;AAED;;;GAGG;AACH,MAAM,MAAM,aAAa,GACrB,mBAAmB,GACnB,mBAAmB,GACnB,2BAA2B,GAC3B,0BAA0B,GAC1B,yBAAyB,GACzB,sBAAsB,GACtB,UAAU,CAAC;AAEf;;GAEG;AACH,MAAM,WAAW,mBAAmB;IAClC,IAAI,EAAE,iBAAiB,CAAC;IACxB,OAAO,EAAE;QACP,EAAE,EAAE,MAAM,CAAC;QACX,WAAW,EAAE,WAAW,CAAC;KAC1B,CAAC;CACH;AAED;;GAEG;AACH,MAAM,WAAW,mBAAmB;IAClC,IAAI,EAAE,iBAAiB,CAAC;IACxB,OAAO,EAAE;QACP,WAAW,EAAE,WAAW,CAAC;KAC1B,CAAC;CACH;AAED;;;GAGG;AACH,MAAM,WAAW,2BAA2B;IAC1C,IAAI,EAAE,0BAA0B,CAAC;IACjC,IAAI,EAAE,MAAM,CAAC;CACd;AAED;;GAEG;AACH,MAAM,WAAW,0BAA0B;IACzC,IAAI,EAAE,wBAAwB,CAAC;IAC/B,aAAa,EAAE,MAAM,CAAC;CACvB;AAED;;;GAGG;AACH,MAAM,WAAW,yBAAyB;IACxC,IAAI,EAAE,uBAAuB,CAAC;IAC9B,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,GAAG,CAAC,EAAE,MAAM,CAAC;IACb,IAAI,EAAE,MAAM,CAAC;IACb,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB;AAED;;;GAGG;AACH,MAAM,WAAW,sBAAsB;IACrC,IAAI,EAAE,oBAAoB,CAAC;IAC3B,IAAI,EAAE,MAAM,CAAC;IACb,QAAQ,CAAC,EAAE,MAAM,CAAC;CACnB;AAED;;GAEG;AACH,MAAM,WAAW,UAAU;IACzB,IAAI,EAAE,OAAO,CAAC;IACd,KAAK,EAAE;QACL,OAAO,EAAE,MAAM,GAAG,OAAO,CAAC;QAC1B,IAAI,CAAC,EAAE,MAAM,CAAC;KACf,CAAC;CACH;AAMD;;GAEG;AACH,MAAM,WAAW,mBAAmB;IAClC;;;;;;;;;;;;;;;;;;;;OAoBG;IACH,UAAU,CACR,WAAW,EAAE,aAAa,CAAC,UAAU,CAAC,EACtC,OAAO,CAAC,EAAE,iBAAiB,GAC1B,aAAa,CAAC,aAAa,CAAC,CAAC;CACjC;AAED;;;;;;;;;;;;;;GAcG;AACH,wBAAgB,yBAAyB,CACvC,MAAM,EAAE,cAAc,GACrB,mBAAmB,CAgFrB;AA0SD;;GAEG;AACH,MAAM,WAAW,kBAAkB;IACjC,2CAA2C;IAC3C,MAAM,EAAE,cAAc,CAAC,UAAU,EAAE,IAAI,EAAE,OAAO,CAAC,CAAC;IAClD,qCAAqC;IACrC,IAAI,EAAE,MAAM,IAAI,CAAC;CAClB;AAED;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA4BG;AACH,wBAAsB,0BAA0B,CAC9C,UAAU,GAAE,MAAc,GACzB,OAAO,CAAC,kBAAkB,CAAC,CAgE7B;AAED;;;;;;;;;;;;;;;;;GAiBG;AACH,wBAAsB,uBAAuB,CAC3C,UAAU,GAAE,MAAc,GACzB,OAAO,CAAC,kBAAkB,CAAC,CAiG7B"}

package/dist/utils/transcription/runtime.d.ts

@@ -0,0 +1,15 @@
+/**
+ * Runtime environment detection for WebSocket implementations
+ *
+ * Provides environment-aware WebSocket constructor selection:
+ * - Browser/Deno: Uses global WebSocket API
+ * - Node.js: Dynamically imports 'ws' package
+ */
+/**
+ * Get the appropriate WebSocket implementation for the current runtime
+ *
+ * @returns WebSocket constructor (browser WebSocket or ws package)
+ * @throws Error if WebSocket is not available in any form
+ */
+export declare function getWebSocketImpl(): Promise<any>;
+//# sourceMappingURL=runtime.d.ts.map

package/dist/utils/transcription/runtime.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"runtime.d.ts","sourceRoot":"","sources":["../../../src/utils/transcription/runtime.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AAEH;;;;;GAKG;AACH,wBAAsB,gBAAgB,IAAI,OAAO,CAAC,GAAG,CAAC,CAyBrD"}

package/dist/utils/transcription/transcribe.d.ts

@@ -25,7 +25,7 @@ export interface TranscribeOptions {
     model?: string;
     /** Enable speaker diarization (default: true) */
     diarize?: boolean;
-    /** Timestamp granularity: 'word' | 'segment' (default: 'word', disabled if language set) */
+    /** Timestamp granularity: 'word' | 'segment' (default: 'segment' when diarize=true, disabled if language set) */
     timestamps?: 'word' | 'segment';
     /** Auto-split long audio (default: true). For URLs, detects duration first. */
     autoSplit?: boolean;

package/dist/utils/transcription/transcribe.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"transcribe.d.ts","sourceRoot":"","sources":["../../../src/utils/transcription/transcribe.ts"],"names":[],"mappings":"AAAA;;GAEG;AASH,OAAO,EAA6B,KAAK,yBAAyB,EAAE,MAAM,2BAA2B,CAAA;AAErG,kDAAkD;AAClD,MAAM,WAAW,gBAAgB;IAC/B,IAAI,EAAE,CAAC,GAAG,EAAE,MAAM,KAAK,IAAI,CAAA;IAC3B,IAAI,EAAE,CAAC,GAAG,EAAE,MAAM,KAAK,IAAI,CAAA;IAC3B,KAAK,EAAE,CAAC,GAAG,EAAE,MAAM,KAAK,IAAI,CAAA;CAC7B;AAQD,MAAM,WAAW,iBAAiB;IAChC,yBAAyB;IACzB,SAAS,CAAC,EAAE,MAAM,CAAA;IAClB,iCAAiC;IACjC,WAAW,CAAC,EAAE,MAAM,CAAA;IACpB,gEAAgE;IAChE,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,iDAAiD;IACjD,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,yEAAyE;IACzE,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,mFAAmF;IACnF,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,kDAAkD;IAClD,KAAK,CAAC,EAAE,MAAM,CAAA;IACd,iDAAiD;IACjD,OAAO,CAAC,EAAE,OAAO,CAAA;IACjB,4FAA4F;IAC5F,UAAU,CAAC,EAAE,MAAM,GAAG,SAAS,CAAA;IAC/B,+EAA+E;IAC/E,SAAS,CAAC,EAAE,OAAO,CAAA;IACnB,iEAAiE;IACjE,cAAc,CAAC,EAAE,MAAM,CAAA;IACvB,uCAAuC;IACvC,MAAM,CAAC,EAAE,gBAAgB,CAAA;IACzB,mCAAmC;IACnC,OAAO,CAAC,EAAE,OAAO,CAAA;CAClB;AAED,MAAM,WAAW,wBAAwB;IACvC,sBAAsB;IACtB,MAAM,EAAE,MAAM,CAAA;IACd,mDAAmD;IACnD,KAAK,CAAC,EAAE,MAAM,CAAA;CACf;AA6FD;;;;;;;;;;;;;;;;;;;GAmBG;AACH,iEAAiE;AACjE,MAAM,WAAW,kBAAkB;IACjC,UAAU,CAAC,OAAO,EAAE,iBAAiB,GAAG,OAAO,CAAC,yBAAyB,CAAC,CAAA;CAC3E;AAED,wBAAgB,wBAAwB,CAAC,MAAM,EAAE,wBAAwB,GAAG,kBAAkB,CA4K7F;AAED,+BAA+B;AAC/B,eAAO,MAAM,UAAU,iCAA2B,CAAA"}
+{"version":3,"file":"transcribe.d.ts","sourceRoot":"","sources":["../../../src/utils/transcription/transcribe.ts"],"names":[],"mappings":"AAAA;;GAEG;AASH,OAAO,EAA6B,KAAK,yBAAyB,EAAE,MAAM,2BAA2B,CAAA;AAErG,kDAAkD;AAClD,MAAM,WAAW,gBAAgB;IAC/B,IAAI,EAAE,CAAC,GAAG,EAAE,MAAM,KAAK,IAAI,CAAA;IAC3B,IAAI,EAAE,CAAC,GAAG,EAAE,MAAM,KAAK,IAAI,CAAA;IAC3B,KAAK,EAAE,CAAC,GAAG,EAAE,MAAM,KAAK,IAAI,CAAA;CAC7B;AAQD,MAAM,WAAW,iBAAiB;IAChC,yBAAyB;IACzB,SAAS,CAAC,EAAE,MAAM,CAAA;IAClB,iCAAiC;IACjC,WAAW,CAAC,EAAE,MAAM,CAAA;IACpB,gEAAgE;IAChE,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,iDAAiD;IACjD,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,yEAAyE;IACzE,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,mFAAmF;IACnF,QAAQ,CAAC,EAAE,MAAM,CAAA;IACjB,kDAAkD;IAClD,KAAK,CAAC,EAAE,MAAM,CAAA;IACd,iDAAiD;IACjD,OAAO,CAAC,EAAE,OAAO,CAAA;IACjB,iHAAiH;IACjH,UAAU,CAAC,EAAE,MAAM,GAAG,SAAS,CAAA;IAC/B,+EAA+E;IAC/E,SAAS,CAAC,EAAE,OAAO,CAAA;IACnB,iEAAiE;IACjE,cAAc,CAAC,EAAE,MAAM,CAAA;IACvB,uCAAuC;IACvC,MAAM,CAAC,EAAE,gBAAgB,CAAA;IACzB,mCAAmC;IACnC,OAAO,CAAC,EAAE,OAAO,CAAA;CAClB;AAED,MAAM,WAAW,wBAAwB;IACvC,sBAAsB;IACtB,MAAM,EAAE,MAAM,CAAA;IACd,mDAAmD;IACnD,KAAK,CAAC,EAAE,MAAM,CAAA;CACf;AA6FD;;;;;;;;;;;;;;;;;;;GAmBG;AACH,iEAAiE;AACjE,MAAM,WAAW,kBAAkB;IACjC,UAAU,CAAC,OAAO,EAAE,iBAAiB,GAAG,OAAO,CAAC,yBAAyB,CAAC,CAAA;CAC3E;AAED,wBAAgB,wBAAwB,CAAC,MAAM,EAAE,wBAAwB,GAAG,kBAAkB,CA4K7F;AAED,+BAA+B;AAC/B,eAAO,MAAM,UAAU,iCAA2B,CAAA"}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@wovin/tranz",
-  "version": "0.1.11",
+  "version": "0.1.15",
   "type": "module",
   "description": "Audio transcription library with provider support and auto-splitting",
   "author": "gotjoshua @gotjoshua",
@@ -43,6 +43,9 @@
     "@mistralai/mistralai": "^1.14.0",
     "execa": "^9.6.1"
   },
+  "optionalDependencies": {
+    "ws": "^8.19.0"
+  },
   "devDependencies": {
     "@types/node": "^24.10.1",
     "@types/ws": "^8.5.13",
@@ -51,7 +54,6 @@
     "tsup": "^8.5.0",
     "tsx": "^4.19.2",
     "typescript": "^5.9.3",
-    "ws": "^8.18.0",
     "yargs": "^17.7.2",
     "tsupconfig": "^0.0.0"
   },