Whisp

A lightweight desktop speech-to-text tool that supports both OpenAI's transcription API and local Whisper models. Whisp provides a simple interface for converting speech to text with minimal resource overhead.

Overview

Whisp offers an unobtrusive and customizable way to transcribe your voice into text. It runs in your system tray and is globally available. Press the hotkey to start recording, speak as long as you want, then press the hotkey again to transcribe and automatically paste into any focused input field.

Design principles:

• Reliable: Built to be stable and handle errors gracefully. Resilient in the face of errors. Retry and recovery.

• Lightweight: Resource-efficient, minimal system impact, simple.

Installation

Recommended: Local Whisper (No API Key Required)

The best experience is with local Whisper models - everything runs on your machine with no API key or internet connection needed. Just install and go:

cargo install whisp --features local-whisper
whisp

On first launch, whisp will automatically download the default model (large-v3-turbo-q8_0, ~834 MiB) and start working immediately.

Requirements: CMake must be installed (for building whisper.cpp). On Linux, ALSA, GTK3, and X11 libraries are also required - a shell.nix is provided for Nix users.

On macOS, Metal GPU acceleration is used automatically, with CoreML support for ~3x faster encoding via the Apple Neural Engine.

Alternative: OpenAI API

If you prefer cloud transcription or can't build local whisper:

cargo install whisp

This requires an OpenAI API key configured in whisp.toml.

Configuration

Configuration is managed through a whisp.toml file located in your system's configuration directory. The whisp tray menu has an option to copy the configuration file path to the clipboard.

Example: Local Whisper Backend

When installed with --features local-whisper, whisp works with zero configuration. Optionally customize with:

local_model = "large-v3-turbo-q8_0"

On first launch, the model is automatically downloaded (~834 MiB for large-v3-turbo-q8_0). The large-v3-turbo model offers excellent transcription quality and runs fast on modern hardware, especially MacBooks with Metal/CoreML acceleration.

Quick start: For faster initial setup, use a smaller model like tiny.en-q8_0 (~42 MiB) or base.en-q8_0 (~78 MiB).

Example: OpenAI Backend

openai_key = "your-api-key"
hotkey = "shift+super+Semicolon"
language = "en"
model = "gpt-4o-mini-transcribe"
restore_clipboard = true
auto_paste = true

Configuration Options

The backend default is local when built with --features local-whisper, otherwise openai.

Available Local Models

Models are downloaded from ggerganov/whisper.cpp on HuggingFace. Model names must match exactly as shown below.

Usage

1. Run whisp - it will appear in your system tray
2. Press the hotkey to start recording
3. Speak as long as you want
4. Press the hotkey again to transcribe and paste

Common Use Cases

• AI Coding Agents: Voice dictate prompts to tools like Claude Code. Much faster than typing.

• Messaging: Quickly respond to messages in chat applications like Discord or Slack.

• Document Writing: Speak freely to draft large amounts of text quickly.

CoreML Acceleration (macOS)

On macOS, the local backend automatically uses CoreML to run the encoder on Apple's Neural Engine, providing ~3x faster encoding performance. This is enabled by default.

On first use, whisp downloads a pre-built CoreML encoder model from HuggingFace (15-1200 MiB depending on model size). The first transcription after launch may be slow as macOS compiles the model for the Neural Engine, but subsequent transcriptions are fast. Since whisp is designed to run in the background indefinitely, this one-time warmup cost is negligible in practice.

CoreML encoders work with all GGML models, including quantized variants. The encoder runs on the Neural Engine while the decoder uses the GGML model on CPU/Metal.

To disable CoreML and use Metal-only acceleration:

coreml = false

License

Whisp is licensed under the MIT license.