AI music video production system utilizes distributed processing

A developer has refined a custom AI-driven music video production system designed to streamline the creative process through distributed processing across multiple high-performance machines, including an ASUS Ascent GX10 (DGX Spark compatible), an RTX 5090 system, and an RTX 4090 system.

The workflow integrates several specialized tools: LaViale for automated lip-syncing music video generation and Vox Score Workbench for vocal editing and music production. The system utilizes the DiffSinger AI voice model to allow users to sing using custom voice clones. To improve lyric generation, the developer integrated the local LLM Qwen 2.5 27B to write meaningful lyrics based on specific themes.

Key enhancements include the addition of an interactive AI avatar capable of lip-syncing and acting as a conversational agent using LLMs like Qwen and Gemma. To manage the complexity of coordinating different software sessions across separate machines, the developer implemented a communication mechanism called ‘Postbox,’ which allows different Claude Code sessions to exchange information automatically.