HiMeditator/auto-caption

What it solves

Auto Caption provides a cross-platform solution for generating real-time subtitles from system audio output or microphone input, enabling users to see live transcriptions and translations of audio streams.

How it works

The software consists of a graphical user interface and a modular caption engine. It captures audio from the system and processes it through a selected speech-to-text (STT) engine. For translation, it can integrate with local LLMs via Ollama, cloud-based OpenAI-compatible APIs, or the Google Translate API. The engine can be run as part of the GUI application or as a standalone executable via the terminal.

Who it’s for

It is designed for users on Windows, macOS, and Linux who need real-time accessibility, translation, or transcription of their system audio or microphone input.

Highlights

• Flexible Engine Options: Supports cloud-based models (Alibaba Gummy, GLM-ASR) and local models (Vosk, SOSV/SenseVoice).
• Multiple Translation Paths: Integrates with local Ollama models (recommended <1B parameters), OpenAI-compatible APIs, and Google Translate.
• Customizable Display: Offers extensive styling options for fonts, colors, and backgrounds.
• Data Export: Allows users to view and export caption records in .srt and .json formats.
• Extensible: Users can develop their own caption engines using Python.