video_transcript_whisper 1.0.1

package/README.md

@@ -0,0 +1,92 @@
+# video_transcript_whisper
+
+A CLI tool that transcribes the audio from a video file to text using [whisper.cpp](https://github.com/ggerganov/whisper.cpp) — no cloud API needed, everything runs locally.
+
+## How it works
+
+1. Extracts audio from the input video and converts it to 16 kHz mono WAV (the format whisper.cpp requires) via bundled `ffmpeg`.
+2. On first run, automatically downloads and compiles whisper.cpp and fetches the chosen Whisper model (default: `medium.en`).
+3. Runs transcription locally and prints the result to stdout, or saves it to a file.
+
+## Installation
+
+```bash
+npm install -g video_transcript_whisper
+```
+
+Or run directly with `npx`:
+
+```bash
+npx video_transcript_whisper -i video.mp4
+```
+
+## Usage
+
+```
+video_transcript_whisper [options]
+
+Options:
+  -i, --input-video-path <path>   Path to the video file to transcribe (required)
+  -o, --output-path <path>        Path to save the transcription (default: print to stdout)
+  -v, --verbose                   Enable verbose logging (default: false)
+  -h, --help                      Display help
+```
+
+### Examples
+
+Print transcription to stdout:
+
+```bash
+video_transcript_whisper -i lecture.mp4
+```
+
+Save transcription to a file:
+
+```bash
+video_transcript_whisper -i interview.mp4 -o transcript.txt
+```
+
+Watch download/compilation progress on first run:
+
+```bash
+video_transcript_whisper -i demo.mp4 -v
+```
+
+## First run
+
+The first time the tool runs it will:
+
+1. Compile whisper.cpp v1.5.5 into `./output/whisper.cpp/` (requires a C++ compiler).
+2. Download the `medium.en` Whisper model (~1.5 GB) into the same directory.
+
+Subsequent runs skip both steps and go straight to transcription.
+
+## Supported video/audio formats
+
+Any format supported by ffmpeg (MP4, MOV, MKV, AVI, MP3, WAV, …).
+
+## Requirements
+
+- Node.js 18+
+- A C++ compiler (`gcc` / `clang`) for the one-time whisper.cpp build step
+- `make`
+
+## Development
+
+```bash
+# Install dependencies
+npm install
+
+# Type-check
+npm run typecheck
+
+# Build
+npm run build
+
+# Run from source
+node dist/cli.js -i video.mp4
+```
+
+## License
+
+MIT

package/dist/cli.d.ts

@@ -0,0 +1,3 @@
+#!/usr/bin/env node
+export {};
+//# sourceMappingURL=cli.d.ts.map

package/dist/cli.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"cli.d.ts","sourceRoot":"","sources":["../src/cli.ts"],"names":[],"mappings":""}

package/dist/cli.js

@@ -0,0 +1,49 @@
+#!/usr/bin/env node
+// node imports
+import Path from 'node:path';
+import Fs from 'node:fs';
+// npm imports
+import * as Commander from 'commander';
+// local imports
+import { TranscriptionWordHelper } from './libs/transcription_word.js';
+import { AudioHelper } from './libs/audio_helper.js';
+const __dirname = new URL('.', import.meta.url).pathname;
+async function main() {
+    // parse command line arguments
+    const program = new Commander.Command();
+    program
+        .option('-i, --input-video-path <path>', 'Path to the video file to transcribe')
+        .option('-o, --output-path <path>', 'Path to save the transcribed captions')
+        .option('-v, --verbose', 'Enable verbose logging', false)
+        .parse(process.argv);
+    const options = program.opts();
+    // Extract audio from video and save as .wav file
+    // convert audio to wav 16KHz Mono - the required format for whisper.cpp
+    const extractedAudioPath = Path.join(__dirname, '../output/extracted_audio.wav');
+    await AudioHelper.convertToWav16kMono(options.inputVideoPath, extractedAudioPath);
+    if (options.verbose) {
+        console.log('Transcribing audio to captions...');
+        console.time('transcribeAudioToCaptions');
+    }
+    // transcribe audio to captions
+    const whisperPath = Path.join(__dirname, '../output', 'whisper.cpp');
+    const transcribedText = await TranscriptionWordHelper.transcribeVoice(whisperPath, extractedAudioPath, { verbose: options.verbose });
+    if (options.verbose) {
+        console.timeEnd('transcribeAudioToCaptions');
+    }
+    // print transcription result
+    if (options.outputPath === undefined) {
+        console.log(transcribedText);
+    }
+    else {
+        // save transcription result to file
+        await Fs.promises.writeFile(options.outputPath, transcribedText, 'utf-8');
+    }
+}
+///////////////////////////////////////////////////////////////////////////////
+///////////////////////////////////////////////////////////////////////////////
+//	
+///////////////////////////////////////////////////////////////////////////////
+///////////////////////////////////////////////////////////////////////////////
+void main();
+//# sourceMappingURL=cli.js.map

package/dist/cli.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"cli.js","sourceRoot":"","sources":["../src/cli.ts"],"names":[],"mappings":";AACA,eAAe;AACf,OAAO,IAAI,MAAM,WAAW,CAAC;AAC7B,OAAO,EAAE,MAAM,SAAS,CAAC;AAEzB,cAAc;AACd,OAAO,KAAK,SAAS,MAAM,WAAW,CAAC;AAEvC,gBAAgB;AAChB,OAAO,EAAE,uBAAuB,EAAE,MAAM,8BAA8B,CAAC;AACvE,OAAO,EAAE,WAAW,EAAE,MAAM,wBAAwB,CAAC;AAErD,MAAM,SAAS,GAAG,IAAI,GAAG,CAAC,GAAG,EAAE,MAAM,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,QAAQ,CAAC;AAEzD,KAAK,UAAU,IAAI;IAClB,+BAA+B;IAC/B,MAAM,OAAO,GAAG,IAAI,SAAS,CAAC,OAAO,EAAE,CAAC;IACxC,OAAO;SACL,MAAM,CAAC,+BAA+B,EAAE,sCAAsC,CAAC;SAC/E,MAAM,CAAC,0BAA0B,EAAE,uCAAuC,CAAC;SAC3E,MAAM,CAAC,eAAe,EAAE,wBAAwB,EAAE,KAAK,CAAC;SACxD,KAAK,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC;IAGtB,MAAM,OAAO,GAAG,OAAO,CAAC,IAAI,EAIxB,CAAC;IAKL,iDAAiD;IACjD,wEAAwE;IACxE,MAAM,kBAAkB,GAAG,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,+BAA+B,CAAC,CAAC;IACjF,MAAM,WAAW,CAAC,mBAAmB,CAAC,OAAO,CAAC,cAAc,EAAE,kBAAkB,CAAC,CAAC;IAElF,IAAI,OAAO,CAAC,OAAO,EAAE,CAAC;QACrB,OAAO,CAAC,GAAG,CAAC,mCAAmC,CAAC,CAAC;QACjD,OAAO,CAAC,IAAI,CAAC,2BAA2B,CAAC,CAAC;IAC3C,CAAC;IAED,+BAA+B;IAC/B,MAAM,WAAW,GAAG,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,WAAW,EAAE,aAAa,CAAC,CAAC;IACrE,MAAM,eAAe,GAAG,MAAM,uBAAuB,CAAC,eAAe,CAAC,WAAW,EAAE,kBAAkB,EAAE,EAAE,OAAO,EAAE,OAAO,CAAC,OAAO,EAAE,CAAC,CAAC;IAGrI,IAAI,OAAO,CAAC,OAAO,EAAE,CAAC;QACrB,OAAO,CAAC,OAAO,CAAC,2BAA2B,CAAC,CAAC;IAC9C,CAAC;IAED,6BAA6B;IAC7B,IAAI,OAAO,CAAC,UAAU,KAAK,SAAS,EAAE,CAAC;QACtC,OAAO,CAAC,GAAG,CAAC,eAAe,CAAC,CAAC;IAC9B,CAAC;SAAM,CAAC;QACP,oCAAoC;QACpC,MAAM,EAAE,CAAC,QAAQ,CAAC,SAAS,CAAC,OAAO,CAAC,UAAU,EAAE,eAAe,EAAE,OAAO,CAAC,CAAC;IAC3E,CAAC;AACF,CAAC;AAED,+EAA+E;AAC/E,+EAA+E;AAC/E,GAAG;AACH,+EAA+E;AAC/E,+EAA+E;AAE/E,KAAK,IAAI,EAAE,CAAC"}

package/dist/libs/audio_helper.d.ts

@@ -0,0 +1,23 @@
+/**
+ * Generic helper class for audio/voice generation and processing using the OpenAI API and MusicMetadata library.
+ * - MUST NOT be specific to the dialog video use case, it should be reusable for other use cases as well.
+ */
+export declare class AudioHelper {
+    /**
+     * Convert an audio file to an MP3 file using ffmpeg.
+     *
+     * @param inputAudioPath input audio path readable by 'ffmpeg-static' npm package
+     * @param outputAudioPath output audio path of the .wav
+     * @returns
+     */
+    static convertToMp3(inputAudioPath: string, outputAudioPath: string): Promise<void>;
+    /**
+     * Convert an audio file to a 16KHz mono wav file using ffmpeg. The output file is required for the whisper.cpp transcriber.
+     *
+     * @param inputAudioPath input audio path readable by 'ffmpeg-static' npm package
+     * @param outputAudioPath output audio path of the .wav
+     * @returns
+     */
+    static convertToWav16kMono(inputAudioPath: string, outputAudioPath: string): Promise<void>;
+}
+//# sourceMappingURL=audio_helper.d.ts.map

package/dist/libs/audio_helper.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"audio_helper.d.ts","sourceRoot":"","sources":["../../src/libs/audio_helper.ts"],"names":[],"mappings":"AAMA;;;GAGG;AAGH,qBAAa,WAAW;IAEvB;;;;;;OAMG;WACU,YAAY,CAAC,cAAc,EAAE,MAAM,EAAE,eAAe,EAAE,MAAM;IAczE;;;;;;OAMG;WACU,mBAAmB,CAAC,cAAc,EAAE,MAAM,EAAE,eAAe,EAAE,MAAM;CAchF"}

package/dist/libs/audio_helper.js

@@ -0,0 +1,53 @@
+// node import
+import ChildProcess from "node:child_process";
+// npm import
+import ffmpegStaticPath from 'ffmpeg-static';
+/**
+ * Generic helper class for audio/voice generation and processing using the OpenAI API and MusicMetadata library.
+ * - MUST NOT be specific to the dialog video use case, it should be reusable for other use cases as well.
+ */
+export class AudioHelper {
+    /**
+     * Convert an audio file to an MP3 file using ffmpeg.
+     *
+     * @param inputAudioPath input audio path readable by 'ffmpeg-static' npm package
+     * @param outputAudioPath output audio path of the .wav
+     * @returns
+     */
+    static async convertToMp3(inputAudioPath, outputAudioPath) {
+        const ffmpegPath = ffmpegStaticPath;
+        const command = `${ffmpegPath} -i "${inputAudioPath}" "${outputAudioPath}" -y`;
+        return new Promise((resolve, reject) => {
+            ChildProcess.exec(command, (error, _stdout, _stderr) => {
+                if (error) {
+                    reject(`Error converting audio: ${error.message}`);
+                }
+                else {
+                    resolve();
+                }
+            });
+        });
+    }
+    /**
+     * Convert an audio file to a 16KHz mono wav file using ffmpeg. The output file is required for the whisper.cpp transcriber.
+     *
+     * @param inputAudioPath input audio path readable by 'ffmpeg-static' npm package
+     * @param outputAudioPath output audio path of the .wav
+     * @returns
+     */
+    static async convertToWav16kMono(inputAudioPath, outputAudioPath) {
+        const ffmpegPath = ffmpegStaticPath;
+        const command = `${ffmpegPath} -i "${inputAudioPath}" -ar 16000 -ac 1 -c:a pcm_s16le "${outputAudioPath}" -y`;
+        return new Promise((resolve, reject) => {
+            ChildProcess.exec(command, (error, _stdout, _stderr) => {
+                if (error) {
+                    reject(`Error converting audio: ${error.message}`);
+                }
+                else {
+                    resolve();
+                }
+            });
+        });
+    }
+}
+//# sourceMappingURL=audio_helper.js.map

package/dist/libs/audio_helper.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"audio_helper.js","sourceRoot":"","sources":["../../src/libs/audio_helper.ts"],"names":[],"mappings":"AAAA,cAAc;AACd,OAAO,YAAY,MAAM,oBAAoB,CAAC;AAE9C,aAAa;AACb,OAAO,gBAAgB,MAAM,eAAe,CAAC;AAE7C;;;GAGG;AAGH,MAAM,OAAO,WAAW;IAEvB;;;;;;OAMG;IACH,MAAM,CAAC,KAAK,CAAC,YAAY,CAAC,cAAsB,EAAE,eAAuB;QACxE,MAAM,UAAU,GAAG,gBAAgB,CAAC;QACpC,MAAM,OAAO,GAAG,GAAG,UAAU,QAAQ,cAAc,MAAM,eAAe,MAAM,CAAC;QAC/E,OAAO,IAAI,OAAO,CAAO,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE;YAC5C,YAAY,CAAC,IAAI,CAAC,OAAO,EAAE,CAAC,KAAmB,EAAE,OAAe,EAAE,OAAe,EAAE,EAAE;gBACpF,IAAI,KAAK,EAAE,CAAC;oBACX,MAAM,CAAC,2BAA2B,KAAK,CAAC,OAAO,EAAE,CAAC,CAAC;gBACpD,CAAC;qBAAM,CAAC;oBACP,OAAO,EAAE,CAAC;gBACX,CAAC;YACF,CAAC,CAAC,CAAC;QACJ,CAAC,CAAC,CAAC;IACJ,CAAC;IAED;;;;;;OAMG;IACH,MAAM,CAAC,KAAK,CAAC,mBAAmB,CAAC,cAAsB,EAAE,eAAuB;QAC/E,MAAM,UAAU,GAAG,gBAAgB,CAAC;QACpC,MAAM,OAAO,GAAG,GAAG,UAAU,QAAQ,cAAc,qCAAqC,eAAe,MAAM,CAAC;QAC9G,OAAO,IAAI,OAAO,CAAO,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE;YAC5C,YAAY,CAAC,IAAI,CAAC,OAAO,EAAE,CAAC,KAAmB,EAAE,OAAe,EAAE,OAAe,EAAE,EAAE;gBACpF,IAAI,KAAK,EAAE,CAAC;oBACX,MAAM,CAAC,2BAA2B,KAAK,CAAC,OAAO,EAAE,CAAC,CAAC;gBACpD,CAAC;qBAAM,CAAC;oBACP,OAAO,EAAE,CAAC;gBACX,CAAC;YACF,CAAC,CAAC,CAAC;QACJ,CAAC,CAAC,CAAC;IACJ,CAAC;CAED"}

package/dist/libs/transcription_word.d.ts

@@ -0,0 +1,32 @@
+import { Language } from '@remotion/install-whisper-cpp';
+/**
+ * from https://www.remotion.dev/docs/install-whisper-cpp/
+ * from https://huggingface.co/openai/whisper-medium
+ */
+export declare class TranscriptionWordHelper {
+    /**
+     * Transcribe an audio file to captions using whisper.cpp. The audio file MUST be a wav file in 16KHz PCM Mono format. You can use
+     * the convertToWav16kMono function in this class to convert an audio file to the required format.
+     *
+     * **NOTE: This function may be long as it may need to install whisper.cpp and download the whisper model the
+     * first time it is run.**
+     *
+     * @param whisperPath path to the whisper.cpp installation directory. This is where the whisper.cpp binary and models will be installed.
+     * @param audioWavPath path to .wav file - the file MUST be a wav file in 16KHz PCM Mono
+     * @param options.verbose if true, will print the output of the various steps (installing whisper.cpp, downloading model, transcribing). Default is false.
+     * @param options.modelName the whisper model to use. Default is 'medium.en'. See https://www.remotion.dev/docs/install-whisper-cpp/#models for available models. "medium.en" is a good default for English audio. If you are transcribing non-English audio, you may want to use "medium" or a model without the ".en" suffix.
+     * @param options.language the language of the audio. This is used to help whisper.cpp transcribe better. Default is 'en' (English). See https://www.remotion.dev/docs/install-whisper-cpp/#languages for available languages
+     */
+    static transcribeVoice(whisperPath: string, audioWavPath: string, { verbose, modelName, language }?: {
+        verbose?: boolean;
+        modelName?: "medium.en" | "base" | "base.en" | "large-v1" | "large-v2" | "large-v3" | "large-v3-turbo" | "medium" | "small" | "small.en" | "tiny" | "tiny.en";
+        language?: Language;
+    }): Promise<string>;
+    /**
+     * Install whisper.cpp and the whisper model if they are not already installed.
+     * @param whisperPath path to the whisper.cpp installation directory. This is where the whisper.cpp binary and models will be installed.
+     * @param verbose
+     */
+    private static _installIfNeeded;
+}
+//# sourceMappingURL=transcription_word.d.ts.map

package/dist/libs/transcription_word.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"transcription_word.d.ts","sourceRoot":"","sources":["../../src/libs/transcription_word.ts"],"names":[],"mappings":"AACA,OAAO,EAAmE,QAAQ,EAAE,MAAM,+BAA+B,CAAC;AAS1H;;;GAGG;AACH,qBAAa,uBAAuB;IAEnC;;;;;;;;;;;;OAYG;WACU,eAAe,CAAC,WAAW,EAAE,MAAM,EAAE,YAAY,EAAE,MAAM,EAAE,EACvE,OAAe,EACf,SAAuB,EACvB,QAAe,EACf,GAAE;QACF,OAAO,CAAC,EAAE,OAAO,CAAC;QAClB,SAAS,CAAC,EAAE,WAAW,GAAG,MAAM,GAAG,SAAS,GAAG,UAAU,GAAG,UAAU,GAAG,UAAU,GAAG,gBAAgB,GAAG,QAAQ,GAAG,OAAO,GAAG,UAAU,GAAG,MAAM,GAAG,SAAS,CAAC;QAC9J,QAAQ,CAAC,EAAE,QAAQ,CAAC;KACf,GAAG,OAAO,CAAC,MAAM,CAAC;IAoCxB;;;;OAIG;mBACkB,gBAAgB;CAqBrC"}

package/dist/libs/transcription_word.js

@@ -0,0 +1,75 @@
+// npm import
+import { downloadWhisperModel, installWhisperCpp, transcribe, toCaptions } from '@remotion/install-whisper-cpp';
+///////////////////////////////////////////////////////////////////////////////
+///////////////////////////////////////////////////////////////////////////////
+//	Class
+///////////////////////////////////////////////////////////////////////////////
+///////////////////////////////////////////////////////////////////////////////
+/**
+ * from https://www.remotion.dev/docs/install-whisper-cpp/
+ * from https://huggingface.co/openai/whisper-medium
+ */
+export class TranscriptionWordHelper {
+    /**
+     * Transcribe an audio file to captions using whisper.cpp. The audio file MUST be a wav file in 16KHz PCM Mono format. You can use
+     * the convertToWav16kMono function in this class to convert an audio file to the required format.
+     *
+     * **NOTE: This function may be long as it may need to install whisper.cpp and download the whisper model the
+     * first time it is run.**
+     *
+     * @param whisperPath path to the whisper.cpp installation directory. This is where the whisper.cpp binary and models will be installed.
+     * @param audioWavPath path to .wav file - the file MUST be a wav file in 16KHz PCM Mono
+     * @param options.verbose if true, will print the output of the various steps (installing whisper.cpp, downloading model, transcribing). Default is false.
+     * @param options.modelName the whisper model to use. Default is 'medium.en'. See https://www.remotion.dev/docs/install-whisper-cpp/#models for available models. "medium.en" is a good default for English audio. If you are transcribing non-English audio, you may want to use "medium" or a model without the ".en" suffix.
+     * @param options.language the language of the audio. This is used to help whisper.cpp transcribe better. Default is 'en' (English). See https://www.remotion.dev/docs/install-whisper-cpp/#languages for available languages
+     */
+    static async transcribeVoice(whisperPath, audioWavPath, { verbose = false, modelName = 'medium.en', language = 'en' } = {}) {
+        // install if needed
+        await this._installIfNeeded(whisperPath, { verbose, modelName });
+        // transcribe
+        const whisperCppOutput = await transcribe({
+            model: modelName,
+            whisperPath: whisperPath,
+            whisperCppVersion: '1.5.5',
+            inputPath: audioWavPath,
+            language: language,
+            tokenLevelTimestamps: true,
+            splitOnWord: false,
+            printOutput: verbose ? true : false,
+        });
+        // Optional: Apply remotion recommended postprocessing
+        const { captions } = toCaptions({
+            whisperCppOutput,
+        });
+        let transcribedText = '';
+        for (const caption of captions) {
+            transcribedText = transcribedText.concat(caption.text);
+        }
+        return transcribedText;
+    }
+    ///////////////////////////////////////////////////////////////////////////////
+    ///////////////////////////////////////////////////////////////////////////////
+    //	Private functions
+    ///////////////////////////////////////////////////////////////////////////////
+    ///////////////////////////////////////////////////////////////////////////////
+    /**
+     * Install whisper.cpp and the whisper model if they are not already installed.
+     * @param whisperPath path to the whisper.cpp installation directory. This is where the whisper.cpp binary and models will be installed.
+     * @param verbose
+     */
+    static async _installIfNeeded(whisperPath, { verbose = false, modelName = 'medium.en', } = {}) {
+        // Install whisper.cpp if needed
+        const whisperAlreadyExisted = await installWhisperCpp({
+            to: whisperPath,
+            version: '1.5.5',
+            printOutput: verbose ? true : false,
+        });
+        // Download the whisper model if needed
+        const modelAlreadyExisted = await downloadWhisperModel({
+            model: modelName,
+            folder: whisperPath,
+            printOutput: verbose ? true : false,
+        });
+    }
+}
+//# sourceMappingURL=transcription_word.js.map

package/dist/libs/transcription_word.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"transcription_word.js","sourceRoot":"","sources":["../../src/libs/transcription_word.ts"],"names":[],"mappings":"AAAA,aAAa;AACb,OAAO,EAAE,oBAAoB,EAAE,iBAAiB,EAAE,UAAU,EAAE,UAAU,EAAY,MAAM,+BAA+B,CAAC;AAG1H,+EAA+E;AAC/E,+EAA+E;AAC/E,QAAQ;AACR,+EAA+E;AAC/E,+EAA+E;AAE/E;;;GAGG;AACH,MAAM,OAAO,uBAAuB;IAEnC;;;;;;;;;;;;OAYG;IACH,MAAM,CAAC,KAAK,CAAC,eAAe,CAAC,WAAmB,EAAE,YAAoB,EAAE,EACvE,OAAO,GAAG,KAAK,EACf,SAAS,GAAG,WAAW,EACvB,QAAQ,GAAG,IAAI,KAKZ,EAAE;QACL,oBAAoB;QACpB,MAAM,IAAI,CAAC,gBAAgB,CAAC,WAAW,EAAE,EAAE,OAAO,EAAE,SAAS,EAAE,CAAC,CAAC;QAEjE,aAAa;QACb,MAAM,gBAAgB,GAAG,MAAM,UAAU,CAAC;YACzC,KAAK,EAAE,SAAS;YAChB,WAAW,EAAE,WAAW;YACxB,iBAAiB,EAAE,OAAO;YAC1B,SAAS,EAAE,YAAY;YACvB,QAAQ,EAAE,QAAQ;YAClB,oBAAoB,EAAE,IAAI;YAC1B,WAAW,EAAE,KAAK;YAClB,WAAW,EAAE,OAAO,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,KAAK;SACnC,CAAC,CAAC;QAEH,sDAAsD;QACtD,MAAM,EAAE,QAAQ,EAAE,GAAG,UAAU,CAAC;YAC/B,gBAAgB;SAChB,CAAC,CAAC;QAEH,IAAI,eAAe,GAAW,EAAE,CAAA;QAChC,KAAK,MAAM,OAAO,IAAI,QAAQ,EAAE,CAAC;YAChC,eAAe,GAAG,eAAe,CAAC,MAAM,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC;QACxD,CAAC;QAED,OAAO,eAAe,CAAA;IACvB,CAAC;IAGD,+EAA+E;IAC/E,+EAA+E;IAC/E,oBAAoB;IACpB,+EAA+E;IAC/E,+EAA+E;IAE/E;;;;OAIG;IACK,MAAM,CAAC,KAAK,CAAC,gBAAgB,CAAC,WAAmB,EAAE,EAC1D,OAAO,GAAG,KAAK,EACf,SAAS,GAAG,WAAW,MAIpB,EAAE;QACL,gCAAgC;QAChC,MAAM,qBAAqB,GAAG,MAAM,iBAAiB,CAAC;YACrD,EAAE,EAAE,WAAW;YACf,OAAO,EAAE,OAAO;YAChB,WAAW,EAAE,OAAO,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,KAAK;SACnC,CAAC,CAAC;QAEH,uCAAuC;QACvC,MAAM,mBAAmB,GAAG,MAAM,oBAAoB,CAAC;YACtD,KAAK,EAAE,SAAS;YAChB,MAAM,EAAE,WAAW;YACnB,WAAW,EAAE,OAAO,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,KAAK;SACnC,CAAC,CAAC;IACJ,CAAC;CACD"}

package/output/.gitignore

@@ -0,0 +1,3 @@
+# ignore all files but .gitignore
+*
+!.gitignore

package/package.json

@@ -0,0 +1,28 @@
+{
+    "name": "video_transcript_whisper",
+    "version": "1.0.1",
+    "description": "",
+    "type": "module",
+    "main": "index.js",
+    "bin": {
+        "video_transcript_whisper": "./dist/cli.js"
+    },
+    "scripts": {
+        "build": "tsc",
+        "typecheck": "tsc --noEmit",
+        "publish:all": "npm run build && npm version patch && npm publish --access public",
+        "test": "echo \"Error: no test specified\" && exit 1"
+    },
+    "keywords": [],
+    "author": "",
+    "license": "MIT",
+    "dependencies": {
+        "@remotion/install-whisper-cpp": "^4.0.446",
+        "commander": "^14.0.3",
+        "ffmpeg-static": "^5.3.0",
+        "typescript": "^6.0.2"
+    },
+    "devDependencies": {
+        "@types/node": "^25.5.2"
+    }
+}

package/src/cli.ts

@@ -0,0 +1,68 @@
+#!/usr/bin/env node
+// node imports
+import Path from 'node:path';
+import Fs from 'node:fs';
+
+// npm imports
+import * as Commander from 'commander';
+
+// local imports
+import { TranscriptionWordHelper } from './libs/transcription_word.js';
+import { AudioHelper } from './libs/audio_helper.js';
+
+const __dirname = new URL('.', import.meta.url).pathname;
+
+async function main() {
+	// parse command line arguments
+	const program = new Commander.Command();
+	program
+		.option('-i, --input-video-path <path>', 'Path to the video file to transcribe')
+		.option('-o, --output-path <path>', 'Path to save the transcribed captions')
+		.option('-v, --verbose', 'Enable verbose logging', false)
+		.parse(process.argv);
+
+
+	const options = program.opts<{
+		inputVideoPath: string;
+		outputPath?: string;
+		verbose: boolean;
+	}>();
+
+
+
+
+	// Extract audio from video and save as .wav file
+	// convert audio to wav 16KHz Mono - the required format for whisper.cpp
+	const extractedAudioPath = Path.join(__dirname, '../output/extracted_audio.wav');
+	await AudioHelper.convertToWav16kMono(options.inputVideoPath, extractedAudioPath);
+
+	if (options.verbose) {
+		console.log('Transcribing audio to captions...');
+		console.time('transcribeAudioToCaptions');
+	}
+
+	// transcribe audio to captions
+	const whisperPath = Path.join(__dirname, '../output', 'whisper.cpp');
+	const transcribedText = await TranscriptionWordHelper.transcribeVoice(whisperPath, extractedAudioPath, { verbose: options.verbose });
+
+
+	if (options.verbose) {
+		console.timeEnd('transcribeAudioToCaptions');
+	}
+
+	// print transcription result
+	if (options.outputPath === undefined) {
+		console.log(transcribedText);
+	} else {
+		// save transcription result to file
+		await Fs.promises.writeFile(options.outputPath, transcribedText, 'utf-8');
+	}
+}
+
+///////////////////////////////////////////////////////////////////////////////
+///////////////////////////////////////////////////////////////////////////////
+//	
+///////////////////////////////////////////////////////////////////////////////
+///////////////////////////////////////////////////////////////////////////////
+
+void main();

package/src/libs/audio_helper.ts

@@ -0,0 +1,57 @@
+// node import
+import ChildProcess from "node:child_process";
+
+// npm import
+import ffmpegStaticPath from 'ffmpeg-static';
+
+/**
+ * Generic helper class for audio/voice generation and processing using the OpenAI API and MusicMetadata library.
+ * - MUST NOT be specific to the dialog video use case, it should be reusable for other use cases as well.
+ */
+
+
+export class AudioHelper {
+
+	/**
+	 * Convert an audio file to an MP3 file using ffmpeg.
+	 * 
+	 * @param inputAudioPath input audio path readable by 'ffmpeg-static' npm package
+	 * @param outputAudioPath output audio path of the .wav
+	 * @returns 
+	 */
+	static async convertToMp3(inputAudioPath: string, outputAudioPath: string) {
+		const ffmpegPath = ffmpegStaticPath;
+		const command = `${ffmpegPath} -i "${inputAudioPath}" "${outputAudioPath}" -y`;
+		return new Promise<void>((resolve, reject) => {
+			ChildProcess.exec(command, (error: Error | null, _stdout: string, _stderr: string) => {
+				if (error) {
+					reject(`Error converting audio: ${error.message}`);
+				} else {
+					resolve();
+				}
+			});
+		});
+	}
+
+	/**
+	 * Convert an audio file to a 16KHz mono wav file using ffmpeg. The output file is required for the whisper.cpp transcriber.
+	 *
+	 * @param inputAudioPath input audio path readable by 'ffmpeg-static' npm package
+	 * @param outputAudioPath output audio path of the .wav
+	 * @returns
+	 */
+	static async convertToWav16kMono(inputAudioPath: string, outputAudioPath: string) {
+		const ffmpegPath = ffmpegStaticPath;
+		const command = `${ffmpegPath} -i "${inputAudioPath}" -ar 16000 -ac 1 -c:a pcm_s16le "${outputAudioPath}" -y`;
+		return new Promise<void>((resolve, reject) => {
+			ChildProcess.exec(command, (error: Error | null, _stdout: string, _stderr: string) => {
+				if (error) {
+					reject(`Error converting audio: ${error.message}`);
+				} else {
+					resolve();
+				}
+			});
+		});
+	}
+
+}

package/src/libs/transcription_word.ts

@@ -0,0 +1,100 @@
+// npm import
+import { downloadWhisperModel, installWhisperCpp, transcribe, toCaptions, Language } from '@remotion/install-whisper-cpp';
+
+
+///////////////////////////////////////////////////////////////////////////////
+///////////////////////////////////////////////////////////////////////////////
+//	Class
+///////////////////////////////////////////////////////////////////////////////
+///////////////////////////////////////////////////////////////////////////////
+
+/**
+ * from https://www.remotion.dev/docs/install-whisper-cpp/
+ * from https://huggingface.co/openai/whisper-medium
+ */
+export class TranscriptionWordHelper {
+
+	/**
+	 * Transcribe an audio file to captions using whisper.cpp. The audio file MUST be a wav file in 16KHz PCM Mono format. You can use 
+	 * the convertToWav16kMono function in this class to convert an audio file to the required format.
+	 * 
+	 * **NOTE: This function may be long as it may need to install whisper.cpp and download the whisper model the 
+	 * first time it is run.**
+	 * 
+	 * @param whisperPath path to the whisper.cpp installation directory. This is where the whisper.cpp binary and models will be installed. 
+	 * @param audioWavPath path to .wav file - the file MUST be a wav file in 16KHz PCM Mono
+	 * @param options.verbose if true, will print the output of the various steps (installing whisper.cpp, downloading model, transcribing). Default is false.
+	 * @param options.modelName the whisper model to use. Default is 'medium.en'. See https://www.remotion.dev/docs/install-whisper-cpp/#models for available models. "medium.en" is a good default for English audio. If you are transcribing non-English audio, you may want to use "medium" or a model without the ".en" suffix.
+	 * @param options.language the language of the audio. This is used to help whisper.cpp transcribe better. Default is 'en' (English). See https://www.remotion.dev/docs/install-whisper-cpp/#languages for available languages
+	 */
+	static async transcribeVoice(whisperPath: string, audioWavPath: string, {
+		verbose = false,
+		modelName = 'medium.en',
+		language = 'en'
+	}: {
+		verbose?: boolean;
+		modelName?: "medium.en" | "base" | "base.en" | "large-v1" | "large-v2" | "large-v3" | "large-v3-turbo" | "medium" | "small" | "small.en" | "tiny" | "tiny.en";
+		language?: Language;
+	} = {}): Promise<string> {
+		// install if needed
+		await this._installIfNeeded(whisperPath, { verbose, modelName });
+
+		// transcribe
+		const whisperCppOutput = await transcribe({
+			model: modelName,
+			whisperPath: whisperPath,
+			whisperCppVersion: '1.5.5',
+			inputPath: audioWavPath,
+			language: language,
+			tokenLevelTimestamps: true,
+			splitOnWord: false,
+			printOutput: verbose ? true : false,
+		});
+
+		// Optional: Apply remotion recommended postprocessing
+		const { captions } = toCaptions({
+			whisperCppOutput,
+		});
+
+		let transcribedText: string = ''
+		for (const caption of captions) {
+			transcribedText = transcribedText.concat(caption.text);
+		}
+
+		return transcribedText
+	}
+
+
+	///////////////////////////////////////////////////////////////////////////////
+	///////////////////////////////////////////////////////////////////////////////
+	//	Private functions
+	///////////////////////////////////////////////////////////////////////////////
+	///////////////////////////////////////////////////////////////////////////////
+
+	/**
+	 * Install whisper.cpp and the whisper model if they are not already installed.
+	 * @param whisperPath path to the whisper.cpp installation directory. This is where the whisper.cpp binary and models will be installed.
+	 * @param verbose 
+	 */
+	private static async _installIfNeeded(whisperPath: string, {
+		verbose = false,
+		modelName = 'medium.en',
+	}: {
+		verbose?: boolean;
+		modelName?: "medium.en" | "base" | "base.en" | "large-v1" | "large-v2" | "large-v3" | "large-v3-turbo" | "medium" | "small" | "small.en" | "tiny" | "tiny.en";
+	} = {}) {
+		// Install whisper.cpp if needed
+		const whisperAlreadyExisted = await installWhisperCpp({
+			to: whisperPath,
+			version: '1.5.5',
+			printOutput: verbose ? true : false,
+		});
+
+		// Download the whisper model if needed
+		const modelAlreadyExisted = await downloadWhisperModel({
+			model: modelName,
+			folder: whisperPath,
+			printOutput: verbose ? true : false,
+		});
+	}
+}

package/tsconfig.json

@@ -0,0 +1,30 @@
+{
+    "compilerOptions": {
+        "target": "ES2020",
+        "module": "node16",
+        "lib": [
+            "ES2020",
+            "DOM"
+        ],
+        "types": [
+            "node"
+        ],
+        "outDir": "./dist",
+        "rootDir": "./src",
+        "strict": true,
+        "esModuleInterop": true,
+        "skipLibCheck": true,
+        "forceConsistentCasingInFileNames": true,
+        "resolveJsonModule": true,
+        "declaration": true,
+        "declarationMap": true,
+        "sourceMap": true
+    },
+    "include": [
+        "src/**/*"
+    ],
+    "exclude": [
+        "node_modules",
+        "dist"
+    ]
+}