whisper.rn 0.1.4 → 0.2.0

package/src/index.tsx

@@ -1,4 +1,10 @@
-import { NativeModules, Platform } from 'react-native'
+import {
+  NativeEventEmitter,
+  DeviceEventEmitter,
+  NativeModules,
+  Platform,
+  DeviceEventEmitterStatic,
+} from 'react-native'
 
 const LINKING_ERROR =
   `The package 'whisper.rn' doesn't seem to be linked. Make sure: \n\n${Platform.select({ ios: "- You have run 'pod install'\n", default: '' })
@@ -15,24 +21,58 @@ const RNWhisper = NativeModules.RNWhisper
     },
   )
 
+let EventEmitter: NativeEventEmitter | DeviceEventEmitterStatic
+if (Platform.OS === 'ios') {
+  EventEmitter = new NativeEventEmitter(RNWhisper)
+}
+if (Platform.OS === 'android') {
+  EventEmitter = DeviceEventEmitter
+}
+
+const EVENT_ON_REALTIME_TRANSCRIBE = '@RNWhisper_onRealtimeTranscribe'
+const EVENT_ON_REALTIME_TRANSCRIBE_END = '@RNWhisper_onRealtimeTranscribeEnd'
+
 export type TranscribeOptions = {
+  /** Spoken language (Default: 'auto' for auto-detect) */
   language?: string,
+  /** Translate from source language to english (Default: false) */
   translate?: boolean,
+  /** Number of threads to use during computation (Default: 4) */
   maxThreads?: number,
+  /** Maximum number of text context tokens to store */
   maxContext?: number,
+  /** Maximum segment length in characters */
   maxLen?: number,
+  /** Enable token-level timestamps */
   tokenTimestamps?: boolean,
+  /** Word timestamp probability threshold */
+  wordThold?: number,
+  /** Time offset in milliseconds */
   offset?: number,
+  /** Duration of audio to process in milliseconds */
   duration?: number,
-  wordThold?: number,
+  /** Tnitial decoding temperature */
   temperature?: number,
   temperatureInc?: number,
+  /** Beam size for beam search */
   beamSize?: number,
+  /** Number of best candidates to keep */
   bestOf?: number,
+  /** Speed up audio by x2 (reduced accuracy) */
   speedUp?: boolean,
+  /** Initial Prompt */
   prompt?: string,
 }
 
+export type TranscribeRealtimeOptions = TranscribeOptions & {
+  /**
+   * Realtime record max duration in seconds. 
+   * Due to the whisper.cpp hard constraint - processes the audio in chunks of 30 seconds,
+   * the recommended value will be <= 30 seconds. (Default: 30)
+   */
+  realtimeAudioSec?: number,
+}
+
 export type TranscribeResult = {
   result: string,
   segments: Array<{
@@ -42,6 +82,32 @@ export type TranscribeResult = {
   }>,
 }
 
+export type TranscribeRealtimeEvent = {
+  contextId: number,
+  jobId: number,
+  /** Is capturing audio, when false, the event is the final result */
+  isCapturing: boolean,
+  code: number,
+  processTime: number,
+  recordingTime: number,
+  data?: TranscribeResult,
+  error?: string,
+}
+
+export type TranscribeRealtimeNativeEvent = {
+  contextId: number,
+  jobId: number,
+  payload: {
+    /** Is capturing audio, when false, the event is the final result */
+    isCapturing: boolean,
+    code: number,
+    processTime: number,
+    recordingTime: number,
+    data?: TranscribeResult,
+    error?: string,
+  },
+}
+
 class WhisperContext {
   id: number
 
@@ -49,8 +115,59 @@ class WhisperContext {
     this.id = id
   }
 
-  async transcribe(path: string, options: TranscribeOptions = {}): Promise<TranscribeResult> {
-    return RNWhisper.transcribe(this.id, path, options)
+  /** Transcribe audio file */
+  transcribe(path: string, options: TranscribeOptions = {}): {
+    /** Stop the transcribe */
+    stop: () => void,
+    /** Transcribe result promise */
+    promise: Promise<TranscribeResult>,
+  } {
+    const jobId: number = Math.floor(Math.random() * 10000)
+    return {
+      stop: () => RNWhisper.abortTranscribe(this.id, jobId),
+      promise: RNWhisper.transcribeFile(this.id, jobId, path, options),
+    }
+  }
+
+  /** Transcribe the microphone audio stream, the microphone user permission is required */
+  async transcribeRealtime(options: TranscribeRealtimeOptions = {}): Promise<{
+    /** Stop the realtime transcribe */
+    stop: () => void,
+    /** Subscribe to realtime transcribe events */
+    subscribe: (callback: (event: TranscribeRealtimeEvent) => void) => void,
+  }> {
+    const jobId: number = Math.floor(Math.random() * 10000)
+    await RNWhisper.startRealtimeTranscribe(this.id, jobId, options)
+    let removeTranscribe: () => void
+    let removeEnd: () => void
+    let lastTranscribePayload: TranscribeRealtimeNativeEvent['payload']
+    return {
+      stop: () => RNWhisper.abortTranscribe(this.id, jobId),
+      subscribe: (callback: (event: TranscribeRealtimeEvent) => void) => {
+        const transcribeListener = EventEmitter.addListener(
+          EVENT_ON_REALTIME_TRANSCRIBE,
+          (evt: TranscribeRealtimeNativeEvent) => {
+            const { contextId, payload } = evt
+            if (contextId !== this.id || evt.jobId !== jobId) return
+            lastTranscribePayload = payload
+            callback({ contextId, jobId: evt.jobId, ...payload })
+            if (!payload.isCapturing) removeTranscribe()
+          }
+        )
+        removeTranscribe = transcribeListener.remove
+        const endListener = EventEmitter.addListener(
+          EVENT_ON_REALTIME_TRANSCRIBE_END,
+          (evt: TranscribeRealtimeNativeEvent) => {
+            const { contextId } = evt
+            if (contextId !== this.id || evt.jobId !== jobId) return
+            callback({ contextId, jobId: evt.jobId, ...lastTranscribePayload, isCapturing: false })
+            removeTranscribe?.()
+            removeEnd()
+          }
+        )
+        removeEnd = endListener.remove
+      },
+    }
   }
 
   async release() {

package/whisper-rn.podspec

@@ -1,8 +1,9 @@
 require "json"
 
 package = JSON.parse(File.read(File.join(__dir__, "package.json")))
-base_compiler_flags = '-DGGML_USE_ACCELERATE -Wno-shorten-64-to-32'
-folly_compiler_flags = '-DFOLLY_NO_CONFIG -DFOLLY_MOBILE=1 -DFOLLY_USE_LIBCPP=1 -Wno-comma -Wno-shorten-64-to-32'
+base_compiler_flags = "-DGGML_USE_ACCELERATE -Wno-shorten-64-to-32"
+folly_compiler_flags = "-DFOLLY_NO_CONFIG -DFOLLY_MOBILE=1 -DFOLLY_USE_LIBCPP=1 -Wno-comma"
+base_optimizer_flags = "-O3 -DNDEBUG"
 
 Pod::Spec.new do |s|
   s.name         = "whisper-rn"
@@ -20,17 +21,23 @@ Pod::Spec.new do |s|
   s.dependency "React-Core"
 
   s.compiler_flags = base_compiler_flags
-  s.xcconfig = {
-    'OTHER_LDFLAGS' => '-framework Accelerate'
+  s.pod_target_xcconfig = {
+    "OTHER_LDFLAGS" => "-framework Accelerate",
+    "OTHER_CFLAGS[config=Release]" => base_optimizer_flags,
+    "OTHER_CPLUSPLUSFLAGS[config=Release]" => base_optimizer_flags
   }
 
   # Don't install the dependencies when we run `pod install` in the old architecture.
   if ENV['RCT_NEW_ARCH_ENABLED'] == '1' then
     s.compiler_flags = base_compiler_flags + " " + folly_compiler_flags + " -DRCT_NEW_ARCH_ENABLED=1"
-    s.pod_target_xcconfig    = {
-        "HEADER_SEARCH_PATHS" => "\"$(PODS_ROOT)/boost\"",
-        "OTHER_CPLUSPLUSFLAGS" => "-DFOLLY_NO_CONFIG -DFOLLY_MOBILE=1 -DFOLLY_USE_LIBCPP=1",
-        "CLANG_CXX_LANGUAGE_STANDARD" => "c++17"
+    new_arch_cpp_flags = "-DFOLLY_NO_CONFIG -DFOLLY_MOBILE=1 -DFOLLY_USE_LIBCPP=1"
+    s.pod_target_xcconfig = {
+      "CLANG_CXX_LANGUAGE_STANDARD" => "c++17",
+      "HEADER_SEARCH_PATHS" => "\"$(PODS_ROOT)/boost\"",
+      "OTHER_LDFLAGS" => "-framework Accelerate",
+      "OTHER_CFLAGS[config=Release]" => base_optimizer_flags,
+      "OTHER_CPLUSPLUSFLAGS[config=Debug]" => new_arch_cpp_flags,
+      "OTHER_CPLUSPLUSFLAGS[config=Release]" => new_arch_cpp_flags + " " + base_optimizer_flags
     }
     s.dependency "React-Codegen"
     s.dependency "RCT-Folly"