>_ AI Video Pipeline
TTS, lip-sync, avatar rendering, and post-production — one desktop workflow.
Producing AI-generated video currently means juggling four or five separate tools: a TTS model, a lip-sync system, an avatar renderer, and a post-production step — each with its own setup, file formats, and failure modes.
This pipeline wraps CosyVoice (TTS), MuseTalk (lip-sync), and a custom rendering layer into a single Tauri desktop app backed by a Python inference server. The workflow goes from script to finished video in a single session, with the inference running locally on our dual RTX 3090 setup.
Primary use case is high-volume content production where consistency and throughput matter more than novelty. Currently in internal use; packaging for general release is next.