The developer introduced 'transono', an open-source Rust-based library designed for audio processing and real-time translation. The project includes a utility for creating virtual audio devices and a TUI application that supports bidirectional translation. While initially developed for Linux, the software has been tested on MacOS.
A second, more versatile open-source adapter connects user audio and text inputs to AI translation models via the OpenAI interface. This tool supports Windows, Linux, and MacOS (ARM architecture). It processes speech from a selected source and routes the translated audio and text to configured outputs, including consoles, SRT files, Telegram channels, and WebSockets.
The adapter was deployed at a festival, providing real-time translation on large screens and generating subtitles for YouTube streams. The developer reported that eight hours of continuous translation service cost slightly over $20. Future plans include expanding support for various streaming inputs and outputs.
Regarding AI model usage, the developer noted that OpenAI and Anthropic have implemented restrictions against using their models to create free clones of paid applications. Users have reported that models like Codex and Claude may refuse requests based on safety policies, while ChatGPT may issue bans for reverse engineering. DeepSeek currently does not enforce such restrictions.
