WebAI Studio is an innovative, client-side artificial intelligence suite that brings the power of deep learning directly to your browser. Built with an uncompromising focus on privacy, absolutely zero user data—neither text nor audio—is transmitted to external servers. All inference happens locally on your machine, and thanks to PWA support, the app functions entirely offline after the initial model caching.

Core Features:

• Persian Text-to-Speech (TTS) & Voice Cloning: Powered by an ONNX port of the Pocket-TTS model. This engine handles Persian text normalization, G2P phoneme conversion, and flow-LM latent generation. It features a Mimi encoder that allows you to clone any voice using just a 5-second reference audio sample.

• AI Voice Splitter: Utilizing the htdemucs model via onnxruntime-web, this tool isolates vocals from instrumental tracks with high precision. It leverages WebGPU for hardware acceleration, falling back to multithreaded WASM for broader compatibility.

• Absolute Privacy: 100% in-browser processing guarantees that your sensitive audio files and scripts remain strictly on your device.

• Modern Architecture: Engineered with Vue 3, TypeScript, Vite, and dedicated Web Workers, ensuring a smooth, non-blocking user interface even during heavy DSP workloads.