### Install Capacitor MLKit Whisper Plugin from npm

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/packages/whisper/README.md

Instructions to install the Capacitor MLKit Whisper plugin from npm once it's published, followed by Capacitor project synchronization.

```bash
npm install @capacitor-mlkit/whisper
npx cap sync
```

--------------------------------

### Install Capacitor MLKit Whisper Plugin for Local Development

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/packages/whisper/README.md

Steps to set up the Capacitor MLKit Whisper plugin for local development, including cloning the repository and installing the package from a local path, then syncing with Capacitor.

```bash
# Clone the repository
git clone https://github.com/ceramicwhite/capacitor-mlkit.git

# Install from local path
npm install ./capacitor-mlkit/packages/whisper
npx cap sync
```

--------------------------------

### Install Node.js Dependencies for Capacitor Plugin Development

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/CONTRIBUTING.md

This command installs all necessary Node.js packages and dependencies defined in the `package.json` file for the Capacitor plugin project, preparing the environment for development.

```shell
npm install
```

--------------------------------

### Install Capacitor MLKit Whisper Plugin from GitHub

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/packages/whisper/README.md

Instructions to install the Capacitor MLKit Whisper plugin directly from its GitHub repository, including options for the main repository or the specific whisper package, followed by Capacitor project synchronization.

```bash
# Install directly from GitHub repository
npm install https://github.com/ceramicwhite/capacitor-mlkit.git#main --save
# or install the specific whisper package
npm install git+https://github.com/ceramicwhite/capacitor-mlkit.git#main:packages/whisper

# Sync with your Capacitor project
npx cap sync
```

--------------------------------

### Install SwiftLint for macOS Code Quality

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/CONTRIBUTING.md

This command installs SwiftLint, a static analysis tool for Swift code, using Homebrew on macOS. It ensures consistent code style and quality for the native iOS part of the Capacitor plugin.

```shell
brew install swiftlint
```

--------------------------------

### Perform Advanced Audio Transcription with Word Timestamps

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/packages/whisper/README.md

TypeScript example showing advanced usage of the Capacitor MLKit Whisper plugin, including transcribing an audio file with language specification, word timestamps, and suppression of non-speech tokens, then iterating through segments and words to display detailed timing and confidence.

```typescript
const result = await Whisper.transcribeFile({
  filePath: '/path/to/audio/file.wav',
  language: 'en',
  wordTimestamps: true,
  suppressNonSpeechTokens: true
});

result.segments.forEach(segment => {
  console.log(`[${segment.startTime}s - ${segment.endTime}s]: ${segment.text}`);
  
  segment.words?.forEach(word => {
    console.log(`  "${word.word}" (${word.confidence.toFixed(2)})`);
  });
});
```

--------------------------------

### Download Core ML Whisper Models for iOS

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/packages/whisper/README.md

Commands to download Core ML models from WhisperKit, recommended for optimal performance on iOS, including instructions to unzip the downloaded files.

```bash
# Download Core ML models (better performance on iOS)
# Base English model (~74MB)
curl -L -o openai_whisper-base.en.mlmodelc.zip https://huggingface.co/argmaxinc/whisperkit-coreml/resolve/main/openai_whisper-base.en/openai_whisper-base.en.mlmodelc.zip
unzip openai_whisper-base.en.mlmodelc.zip

# Tiny English model (~39MB) - fastest
# curl -L -o openai_whisper-tiny.en.mlmodelc.zip https://huggingface.co/argmaxinc/whisperkit-coreml/resolve/main/openai_whisper-tiny.en/openai_whisper-tiny.en.mlmodelc.zip

# Small English model (~244MB) - better accuracy
# curl -L -o openai_whisper-small.en.mlmodelc.zip https://huggingface.co/argmaxinc/whisperkit-coreml/resolve/main/openai_whisper-small.en/openai_whisper-small.en.mlmodelc.zip
```

--------------------------------

### Performance Optimization Tips for Whisper Plugin

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/packages/whisper/README.md

Provides practical advice for optimizing the performance of the Whisper plugin, focusing on model selection (Core ML, English-only, smaller models), acceleration techniques, preloading, thread management, and balancing model size with accuracy requirements.

```APIDOC
1. Use Core ML models (.mlmodelc) for maximum performance on iOS
2. Use English-only models when possible for better performance
3. Start with smaller models (tiny, base) and upgrade if needed
4. Enable Core ML acceleration for significant performance improvements
5. Preload models during app initialization for faster transcription
6. Use appropriate thread counts based on device capabilities
7. Consider model size vs accuracy tradeoffs based on your use case
```

--------------------------------

### Download GGML Whisper Models for iOS

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/packages/whisper/README.md

Commands to download standard GGML (whisper.cpp format) models, such as `base.en`, for use with the iOS application bundle.

```bash
# Download a model (example: base.en model ~74MB)
curl -L -o ggml-base.en.bin https://huggingface.co/ggml-org/whisper.cpp/resolve/main/ggml-base.en.bin

# Other available models:
# curl -L -o ggml-tiny.en.bin https://huggingface.co/ggml-org/whisper.cpp/resolve/main/ggml-tiny.en.bin
# curl -L -o ggml-small.en.bin https://huggingface.co/ggml-org/whisper.cpp/resolve/main/ggml-small.en.bin
```

--------------------------------

### Build Capacitor Plugin Web Assets and API Documentation

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/CONTRIBUTING.md

This script compiles TypeScript source code into ESM JavaScript, bundles it for browser usage, and generates API documentation using `@capacitor/docgen`. It prepares the plugin's web assets for use in applications.

```shell
npm run build
```

--------------------------------

### Perform Basic Audio Transcription with Capacitor MLKit Whisper

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/packages/whisper/README.md

TypeScript code demonstrating how to load a Whisper model and transcribe an audio file using the Capacitor MLKit Whisper plugin, then log the transcription text and processing time.

```typescript
import { Whisper } from '@capacitor-mlkit/whisper';

// Load a model
await Whisper.loadModel({
  modelName: 'base.en',
  useCoreML: true
});

// Transcribe an audio file
const result = await Whisper.transcribeFile({
  filePath: '/path/to/audio/file.wav'
});

console.log('Transcription:', result.text);
console.log('Processing time:', result.processingTime, 'ms');
```

--------------------------------

### Plugin License Information

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/packages/whisper/README.md

States the licensing terms for the Capacitor MLKit Whisper plugin, indicating it is distributed under the Apache-2.0 license.

```APIDOC
Apache-2.0
```

--------------------------------

### Supported Whisper Models (GGML & Core ML)

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/packages/whisper/README.md

Lists the various GGML (whisper.cpp) and Core ML models supported by the plugin, detailing their sizes and characteristics (e.g., speed, accuracy, English-only vs. multilingual). It also provides a recommendation to use Core ML models for optimal iOS performance.

```APIDOC
GGML Models (whisper.cpp format):
- tiny (39 MB) - Fastest, lowest accuracy
- tiny.en (39 MB) - English-only, faster than multilingual tiny
- base (74 MB) - Good balance of speed and accuracy
- base.en (74 MB) - English-only, better accuracy than multilingual base
- small (244 MB) - Better accuracy, slower processing
- small.en (244 MB) - English-only small model
- medium (769 MB) - High accuracy, slower processing
- large-v3 (1550 MB) - Highest accuracy, slowest processing

Core ML Models (iOS Optimized):
Available from WhisperKit Core ML Models:
- openai_whisper-tiny.en (39 MB) - Hardware-accelerated tiny model
- openai_whisper-base.en (74 MB) - Hardware-accelerated base model
- openai_whisper-small.en (244 MB) - Hardware-accelerated small model
- openai_whisper-large-v3 (1550 MB) - Hardware-accelerated large model

Recommendation: Use Core ML models (.mlmodelc) for best performance on iOS devices with Apple Neural Engine acceleration.
```

--------------------------------

### LoadModelOptions Interface Reference

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/packages/whisper/README.md

Specifies the parameters for loading a Whisper model, including the model name, whether to utilize Core ML acceleration on iOS, and if the plugin is allowed to download the model if it's not found locally.

```APIDOC
LoadModelOptions:
  modelName: string
    Name of the model to load.
  useCoreML: boolean
    Whether to use Core ML acceleration (iOS only). Default: true
  allowDownload: boolean
    Whether to download the model if not available locally. Default: true
```

--------------------------------

### Whisper Plugin API Methods

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/packages/whisper/README.md

This section outlines the core methods available in the Capacitor MLKit Whisper plugin. It includes functions for transcribing audio files, loading and unloading models, retrieving model information, checking and requesting permissions, and verifying plugin support on the device.

```APIDOC
transcribeFile(options: TranscribeFileOptions) => Promise<TranscribeResult>
  Transcribe audio from a file path.

loadModel(options: LoadModelOptions) => Promise<void>
  Load a Whisper model for transcription.

unloadModel() => Promise<void>
  Unload the currently loaded model to free up memory.

getModelInfo() => Promise<ModelInfo>
  Get information about the currently loaded model.

isSupported() => Promise<IsSupportedResult>
  Check if whisper.cpp is supported on this device.

checkPermissions() => Promise<PermissionStatus>
  Check microphone permission status.

requestPermissions() => Promise<PermissionStatus>
  Request microphone permission.
```

--------------------------------

### Verify Capacitor Plugin Builds Across Platforms

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/CONTRIBUTING.md

This script builds and validates both the web and native projects of the Capacitor plugin. It is primarily used in CI environments to ensure the plugin compiles correctly for all target platforms.

```shell
npm run verify
```

--------------------------------

### Manage Whisper Model Lifecycle in Capacitor

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/packages/whisper/README.md

This snippet demonstrates how to check Whisper.cpp and Core ML support, retrieve information about the currently loaded model (name, Core ML usage, size), and unload the model to free up memory. It covers essential operations for managing the Whisper transcription model within a Capacitor application.

```typescript
// Check if whisper.cpp is supported
const { supported, coreMLSupported } = await Whisper.isSupported();
console.log('Whisper supported:', supported);
console.log('Core ML supported:', coreMLSupported);

// Get model information
const modelInfo = await Whisper.getModelInfo();
console.log('Current model:', modelInfo.modelName);
console.log('Using Core ML:', modelInfo.usingCoreML);
console.log('Model size:', modelInfo.modelSize, 'bytes');

// Unload model to free memory
await Whisper.unloadModel();
```

--------------------------------

### Check and Autoformat Capacitor Plugin Code Quality

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/CONTRIBUTING.md

These scripts check code formatting and quality using ESLint, Prettier, and SwiftLint, and attempt to autoformat or autofix issues. They enforce a consistent style and structure for easier collaboration.

```shell
npm run lint
npm run fmt
```

--------------------------------

### Configure iOS Info.plist for Microphone Access

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/packages/whisper/README.md

XML snippet to add the `NSMicrophoneUsageDescription` key to your iOS `Info.plist` file, which is required for microphone access for audio recording and transcription.

```xml
<key>NSMicrophoneUsageDescription</key>
<string>This app needs access to the microphone to record audio for transcription.</string>
```

--------------------------------

### TranscribeFileOptions Interface Reference

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/packages/whisper/README.md

Defines the options available for the `transcribeFile` method, allowing customization of transcription parameters such as file path, language, word timestamps, translation, thread count, context length, beam search, and non-speech token suppression.

```APIDOC
TranscribeFileOptions:
  filePath: string
    Path to the audio file to transcribe.
  language: string
    Language code for transcription (e.g., 'en', 'es', 'fr').
  wordTimestamps: boolean
    Whether to include word-level timestamps. Default: false
  translate: boolean
    Whether to translate to English. Default: false
  threadCount: number
    Number of threads to use (iOS only).
  maxContext: number
    Maximum context length. Default: 224
  beamSearch: boolean
    Whether to use beam search decoding. Default: false
  suppressNonSpeechTokens: boolean
    Whether to suppress non-speech tokens. Default: true
```

--------------------------------

### Handle Transcription Errors in Capacitor Whisper Plugin

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/packages/whisper/README.md

Demonstrates a `try-catch` block for handling potential errors during audio transcription. It shows how to catch generic errors and specifically check for a 'MODEL_NOT_LOADED' error to attempt model loading and retry the operation, ensuring robust application behavior.

```typescript
try {
  const result = await Whisper.transcribeFile({
    filePath: '/path/to/audio.wav'
  });
} catch (error) {
  console.error('Transcription failed:', error);
  
  // Handle specific error types
  if (error.message.includes('MODEL_NOT_LOADED')) {
    await Whisper.loadModel({ modelName: 'base.en' });
    // Retry transcription
  }
}
```

--------------------------------

### TranscribeResult Interface Reference

Source: https://github.com/ceramicwhite/capacitor-mlkit-whisper/blob/main/packages/whisper/README.md

Describes the structure of the result returned after an audio transcription, including the full transcribed text, an array of segments with timestamps, the detected language code, and the total processing time in milliseconds.

```APIDOC
TranscribeResult:
  text: string
    The transcribed text.
  segments: TranscriptionSegment[]
    Array of transcribed segments with timestamps.
  language: string
    Detected language code.
  processingTime: number
    Total processing time in milliseconds.
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.