Neura Market
abb128/LiveCaptions logo

abb128/LiveCaptions

Free

Linux Desktop application that provides live captioning

FreeFree tier
Type
Open Source
Visit Website GitHub (38 stars)

About abb128/LiveCaptions

Live Captions is an open-source Linux desktop application that provides real-time captioning using on-device speech recognition. It captures desktop audio or microphone input and displays captions locally, ensuring privacy as no audio data is ever sent to external servers. The application features a simple, adjustable interface with options to change font, size, text casing, and optional token-level confidence fading. Built on the aprilasr library and using pretrained models, it requires a reasonably modern CPU (tested on Intel i5/i7, AMD Ryzen 5, and Steam Deck) but does not rely on a GPU. Currently only English is supported; other languages may produce inaccurate results. The project is maintained on GitHub and includes detailed build instructions for GNOME Builder or terminal.

Key Features

Simple user interface for real-time captions
Captures desktop audio or microphone input locally
No audio data sent over the network – 100% on-device
No proprietary services or libraries required
Adjustable font, font size, and text casing
Optional token-level confidence text fading
Built on the aprilasr real-time speech recognition library

Pros & Cons

Pros
  • Fully open-source and free of charge
  • Privacy-focused: all processing happens locally
  • Simple, lightweight interface with customizable display options
  • No dependency on cloud services or proprietary APIs
  • Active development with community support via Discord
  • Works on older CPUs and does not require a GPU
Cons
  • Currently supports English only; other languages may produce gibberish
  • Requires a decent CPU for real-time performance, may impact multitasking
  • Accuracy is not guaranteed; captions may contain errors (especially numbers)
  • Limited to Linux desktop (no Windows/macOS support out of the box)
  • Setup requires building from source or installing dependencies
  • No official packaged release or prebuilt binaries provided

Best For

Providing live captions for video calls and meetings on LinuxAssisting users who are deaf or hard of hearing with system-wide audioCaptioning media playback or online streamsEducational settings for note-taking or accessibilityAccessible computing for hearing-impaired Linux users

FAQ

What languages are supported?
Only the English language is supported currently. Other languages may produce gibberish or a bad phonetic translation.
Is my audio sent to the cloud?
No. Captioning is done entirely on-device; audio is never sent anywhere.
What hardware is required?
A somewhat-decent CPU that can perform realtime captioning is needed. It has been tested on Intel i7-2670QM, i7-7820HQ, i5-8265U, AMD Ryzen 5 1600, and Steam Deck. GPU is not required.
Can I use this on Windows or macOS?
The application is designed for Linux desktop. Build instructions mention macOS (x86) support, but it is primarily Linux-focused.
How accurate are the captions?
The live captions may not be accurate. The tool may make mistakes, including with numbers. It is not recommended for critical or important tasks.
How do I build Live Captions?
You need to clone the repository recursively to get dependencies. Building can be done easily with GNOME Builder or from the terminal with ONNXRuntime v1.14.1 and dependencies like pulseaudio and libadwaita.