#speech-to-text #transcribe #dictation

bin+lib whisp

Unobtrusive global speech-to-text

3 releases

0.1.2 Oct 19, 2024
0.1.1 Oct 19, 2024
0.1.0 Oct 17, 2024

#4 in Accessibility

Download history 438/week @ 2024-10-14 59/week @ 2024-10-21

497 downloads per month

MIT license

46KB
922 lines

Whisp

A lightweight desktop speech-to-text tool powered by modern models like OpenAI's Whisper. Whisp provides a simple interface for converting speech to text with minimal resource overhead.

Crates.io MIT licensed

Overview

Whisp offers an unobtrusive and customizable way to transcribe your voice into text. It operates as a globally available desktop application. Activate it via a hotkey, and it can automatically paste the transcribed text 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

Currently the only way to install this is via cargo:

cargo install whisp
whisp
# if cargo bin is not in your path
~/.cargo/bin/whisp

Configuration

Configuration is managed through a whisp.toml file located in your systems configuration directory. The whisp drop-down has an option to copy the configuration file path to the clipboard.

hotkey = "shift+super+Semicolon"
openai_key = "your-api-key"
language = "en"
model = "whisper-1"
restore_clipboard = true
auto_paste = false

Usage

To start using Whisp, define your preferred hotkey, configure the model, and run the application. You can then trigger voice recording via the hotkey and receive transcriptions automatically.

Common Use Cases

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

  • Document Writing: Speak freely to draft large amounts of text quickly. Then apply post-processing yourself or with the help of a language model to refine the text.

  • Code Commenting: Dictate comments directly into your editor. Note this tool does not write code well. However, perhaps this can change when the automatic post processing is added. Reach out if you are interested in contributing.

License

Whisp is licensed under the MIT license.

Dependencies

~18–58MB
~1M SLoC