whisper-cpp-ts

whisper-cpp-ts is a TypeScript library that wraps the whisper.cpp CLI so you can run local speech-to-text transcription from Node.js.

Requirements

• A compiled whisper.cpp binary placed at bin/whisper (or bin/whisper.exe on Windows).
• A model file such as ggml-base.bin placed at models/ggml-base.bin.

You can download models from the official whisper.cpp releases or build them following the instructions in the upstream repository.

Usage

import { transcribe } from "ts-whisper";

const result = await transcribe({
  file: "audio.wav",
  language: "pt",
});

console.log(result.text);

Options

• modelPath: Override the model file path. Defaults to models/ggml-base.bin.
• threads: Set the number of CPU threads to use with whisper.cpp.
• translate: When true, whisper.cpp will translate speech to English.

Binary and model helpers

• checkWhisperBinary() verifies the binary exists and returns the resolved path.
• checkModel() verifies the model exists and returns the resolved path.

Notes

This library only orchestrates the whisper.cpp CLI and does not reimplement any Whisper inference.