Merge ccbfe358be into 1fe009caed

bindings/node.js/.gitignore

@@ -0,0 +1,2 @@
+/node_modules/
+/whisper.cpp

bindings/node.js/README.md

@@ -0,0 +1,166 @@
+# NPM package wrapping the whisper.cpp Node.js addon
+
+This package is useful in a node.js environment (node, Electron, etc.) where it will provide access to the C++ implementation of whisper.cpp which should be the fastest possible way to run it.
+
+## Install
+To install you need to have [cmake v4+](https://cmake.org/download/) (and [Visual Studio](https://github.com/nodejs/node-gyp#on-windows) on Windows) already available on you machine - it will be used to build the addon for your environment.
+
+```shell
+npm install whisper.cpp.node
+npx whisper.cpp.node install
+```
+
+This will download the whisper.cpp repository and will build the addon. There are few commands that can be used:
+
+- "install" - downloads the latest released tag of the whisper.cpp repository at the time of publishing or does nothing if the repo is already downloaded;
+- "install latest" - downloads the latest master;
+- "reinstall" - downloads the latest released tag of the whisper.cpp repository at the time of publishing even if the repo is already downloaded;
+- "reinstall latest" - will produce the same result as "install latest";
+- "rebuild" - will not download anything but will simply try to rebuild the addon (if for example you change the version of node).
+
+The addon will then be available as the package's main export for use like:
+
+```javascript
+const whisper = require("whisper.cpp.node");
+
+const transcription = await whisper({
+  language: 'en',
+  model: './models/ggml-base.en.bin',
+  fname_inp: './your/file/here'
+});
+
+console.log(transcription);
+```
+
+Check the [Supported Parameters section](#supported-parameters) for more parameter information.
+
+## What the package will not provide for you
+
+It will not download any models for inference. As noted in many other packages, there are models ready for download and use in the [Hugging Face repo of whisper.cpp](https://huggingface.co/ggerganov/whisper.cpp/tree/main).
+
+It will also not work if you try to bundle it for the browser (you should use the [whisper.cpp](https://www.npmjs.com/package/whisper.cpp) package instead which provides the WASM version).
+
+## Links
+
+- [GitHub Repository](https://github.com/gkostov/whisper.cpp.node)
+- [NPM Package](https://www.npmjs.com/package/whisper.cpp.node)
+
+
+And following is the original README of the addon where you can see details for using it.
+______
+
+# whisper.cpp Node.js addon
+
+This is an addon demo that can **perform whisper model reasoning in `node` and `electron` environments**, based on [cmake-js](https://github.com/cmake-js/cmake-js).
+It can be used as a reference for using the whisper.cpp project in other node projects.
+
+This addon now supports **Voice Activity Detection (VAD)** for improved transcription performance.
+
+## Install
+
+```shell
+npm install
+```
+
+## Compile
+
+Make sure it is in the project root directory and compiled with make-js.
+
+```shell
+npx cmake-js compile -T addon.node -B Release
+```
+
+For Electron addon and cmake-js options, you can see [cmake-js](https://github.com/cmake-js/cmake-js) and make very few configuration changes.
+
+> Such as appointing special cmake path:
+> ```shell
+> npx cmake-js compile -c 'xxx/cmake' -T addon.node -B Release
+> ```
+
+## Run
+
+### Basic Usage
+
+```shell
+cd examples/addon.node
+
+node index.js --language='language' --model='model-path' --fname_inp='file-path'
+```
+
+### VAD (Voice Activity Detection) Usage
+
+Run the VAD example with performance comparison:
+
+```shell
+node vad-example.js
+```
+
+## Voice Activity Detection (VAD) Support
+
+VAD can significantly improve transcription performance by only processing speech segments, which is especially beneficial for audio files with long periods of silence.
+
+### VAD Model Setup
+
+Before using VAD, download a VAD model:
+
+```shell
+# From the whisper.cpp root directory
+./models/download-vad-model.sh silero-v6.2.0
+```
+
+### VAD Parameters
+
+All VAD parameters are optional and have sensible defaults:
+
+- `vad`: Enable VAD (default: false)
+- `vad_model`: Path to VAD model file (required when VAD enabled)
+- `vad_threshold`: Speech detection threshold 0.0-1.0 (default: 0.5)
+- `vad_min_speech_duration_ms`: Min speech duration in ms (default: 250)
+- `vad_min_silence_duration_ms`: Min silence duration in ms (default: 100)
+- `vad_max_speech_duration_s`: Max speech duration in seconds (default: FLT_MAX)
+- `vad_speech_pad_ms`: Speech padding in ms (default: 30)
+- `vad_samples_overlap`: Sample overlap 0.0-1.0 (default: 0.1)
+
+### JavaScript API Example
+
+```javascript
+const path = require("path");
+const { whisper } = require(path.join(__dirname, "../../build/Release/addon.node"));
+const { promisify } = require("util");
+
+const whisperAsync = promisify(whisper);
+
+// With VAD enabled
+const vadParams = {
+  language: "en",
+  model: path.join(__dirname, "../../models/ggml-base.en.bin"),
+  fname_inp: path.join(__dirname, "../../samples/jfk.wav"),
+  vad: true,
+  vad_model: path.join(__dirname, "../../models/ggml-silero-v6.2.0.bin"),
+  vad_threshold: 0.5,
+  progress_callback: (progress) => console.log(`Progress: ${progress}%`)
+};
+
+whisperAsync(vadParams).then(result => console.log(result));
+```
+
+## Supported Parameters
+
+Both traditional whisper.cpp parameters and new VAD parameters are supported:
+
+- `language`: Language code (e.g., "en", "es", "fr")
+- `model`: Path to whisper model file
+- `fname_inp`: Path to input audio file
+- `use_gpu`: Enable GPU acceleration (default: true)
+- `flash_attn`: Enable flash attention (default: false)
+- `no_prints`: Disable console output (default: false)
+- `no_timestamps`: Disable timestamps (default: false)
+- `detect_language`: Auto-detect language (default: false)
+- `audio_ctx`: Audio context size (default: 0)
+- `max_len`: Maximum segment length (default: 0)
+- `max_context`: Maximum context size (default: -1)
+- `prompt`: Initial prompt for decoder
+- `comma_in_time`: Use comma in timestamps (default: true)
+- `print_progress`: Print progress info (default: false)
+- `progress_callback`: Progress callback function
+- VAD parameters (see above section)

bindings/node.js/bin/build.js

@@ -0,0 +1,84 @@
+#!/usr/bin/env node
+
+const { execSync } = require('child_process');
+const fs = require('fs');
+const path = require('path');
+const https = require('https');
+const { createWriteStream } = require('fs');
+
+async function downloadAndUnzip(url, filename, output) {
+	const file = createWriteStream(filename);
+
+	console.log(`Downloading from ${url} ...`);
+	await new Promise((resolve, reject) => {
+		https.get(url, (response) => {
+			if (response.statusCode === 302 && response.headers.location) {
+				https.get(response.headers.location, (redirectResponse) => {
+					redirectResponse.pipe(file);
+					file.on('finish', resolve);
+					file.on('error', reject);
+				}).on('error', reject);
+			} else {
+				response.pipe(file);
+				file.on('finish', resolve);
+				file.on('error', reject);
+			}
+		}).on('error', reject);
+	});
+	console.log(`Download complete. Extracting...`);
+
+	const unzipCmd = process.platform === 'win32' ? 
+		`powershell -command "Expand-Archive -Path '${filename}' -DestinationPath '${output}' -Force"` :
+		`unzip -o "${filename}" -d "${output}"`;
+	execSync(unzipCmd, { stdio: 'inherit' });
+	fs.unlinkSync(filename);
+}
+
+const isLatestRequested = process.argv[3] == 'latest';
+const releaseUrl = isLatestRequested ? 'https://github.com/ggml-org/whisper.cpp/archive/refs/heads/master.zip' : 'https://github.com/ggml-org/whisper.cpp/archive/refs/tags/v1.8.3.zip';
+const whisperCppDir = path.resolve(__dirname, '..', 'whisper.cpp');
+const buildCmd = `cd ${whisperCppDir} && npx cmake-js clean && npx cmake-js compile -T addon.node -B Release`;
+
+async function install() {
+	console.log(
+		'whisper.cpp will now be downloaded from the github repo and the addon will be built.'
+		+ '\nIt is going to a take a while ...'
+	);
+	// extract to a .tmp directory because it's not yet know what directory the files will be extracted to
+	const whisperCppTmpDir = whisperCppDir + '.tmp';
+	await downloadAndUnzip(releaseUrl, path.resolve(__dirname, '..', releaseUrl.split('/').pop()), whisperCppTmpDir);
+	// move the extracted files out from the first (and only) directory inside to "whisper.cpp"
+	fs.renameSync(path.resolve(whisperCppTmpDir, fs.readdirSync(whisperCppTmpDir)[0]), whisperCppDir);
+	fs.rmdirSync(whisperCppTmpDir);
+	console.log(`Building ...`);
+	execSync(buildCmd, { stdio: 'inherit' });
+	console.log('whisper.cpp.node should now be ready for use.');
+}
+
+if (process.argv[2] == 'install') {
+	if (isLatestRequested) {
+		if (fs.existsSync(whisperCppDir)){
+			console.log('Replacing with the latest one from master.');
+			execSync(`rm -rf ${whisperCppDir}`, { stdio: 'inherit' });
+		}
+		install();
+	} else if (fs.existsSync(whisperCppDir)) {
+		console.log(
+			'The directory with whisper.cpp exists so the addon should be already available for use.'
+			+ '\nIf you think there\'s something wrong with the installation and would like to install afresh, then run again with the "reinstall" option.'
+		);
+	} else
+		install();
+} else if (process.argv[2] == 'reinstall') {
+	// remove if already there and install
+	execSync(`rm -rf ${whisperCppDir}`, { stdio: 'inherit' });
+	install();
+} else if (process.argv[2] == 'rebuild') {
+	console.log(
+		'whisper.cpp.node will now be rebuilt.'
+		+ '\nIt is going to a take a while ...'
+	);
+	execSync(buildCmd, { stdio: 'inherit' });
+	console.log('whisper.cpp.node should now be ready for use.');
+} else
+	console.log('Not sure what you want to do right now. Please pick one of "install", "reinstall" or "rebuild" commands.');

bindings/node.js/package.json

@@ -0,0 +1,35 @@
+{
+  "name": "whisper.cpp.node",
+  "version": "1.8.3-1",
+  "description": "NPM package wrapping the whisper.cpp addon for Node.js",
+  "main": "whisper.js",
+  "scripts": {
+    "test": "echo \"Error: no test specified\" && exit 1"
+  },
+  "repository": {
+    "type": "git",
+    "url": "git+https://github.com/ggerganov/whisper.cpp"
+  },
+  "keywords": [
+    "openai",
+    "whisper",
+    "speech-to-text",
+    "speech-recognition",
+    "transformer"
+  ],
+  "bin": "bin/build.js",
+  "files": [
+    "index.js",
+    "bin/"
+  ],
+  "author": "Georgi Kostov",
+  "license": "MIT",
+  "dependencies": {
+    "cmake-js": "^7.4.0",
+    "node-addon-api": "^5.0.0"
+  },
+  "bugs": {
+    "url": "https://github.com/ggerganov/whisper.cpp/issues"
+  },
+  "homepage": "https://github.com/ggerganov/whisper.cpp#readme"
+}

bindings/node.js/whisper.js

@@ -0,0 +1,9 @@
+const path = require('path');
+const os = require('os');
+const { promisify } = require('util');
+
+const buildPath = os.platform() == 'win32' ? './whisper.cpp/build/bin/Release/addon.node' : './whisper.cpp/build/Release/addon.node';
+
+const { whisper } = require(path.join(__dirname, buildPath));
+
+module.exports = promisify(whisper);