Sharrnah/whispering-ui

What it solves

Whispering Tiger UI provides a user-friendly desktop interface for local AI-powered audio transcription, translation, and text-to-speech. It eliminates the need for command-line tools by allowing users to process audio streams (microphone or desktop audio) and images (OCR) locally and offline, with the integration of external outputs like VRChat via OSC or browser overlays via WebSockets.

How it works

The project consists of a native Go/Fyne UI that connects to a Python backend. It supports various compute devices, including CPU, NVIDIA CUDA, and Vulkan (via audio.cpp) for AMD and Intel hardware. Users can create profiles to configure audio input/output devices, select AI models and their precision, and manage plugins for extended functionality.

Who it’s for

Users on Windows and Linux who need real-time translation and transcription of their own voice or system audio, as well as those who want to integrate AI audio tools into virtual environments like VRChat.

Highlights

• Local Processing: AI models run locally and offline after the initial download.
• Multi-Hardware Support: Compatible with NVIDIA CUDA, AMD, Intel, and Vulkan.
• Versatile Audio Capture: Supports WASAPI loopback (Windows) and PulseAudio/PipeWire (Linux) for desktop audio capture.
• Extensible: Includes a plugin system for features like Emotion Prediction, RVC (Retrieval-based Voice Conversion), and LLM integration.
• Integrated OCR: Capable of recognizing and translating text from in-game images.