🎛️ Audio-Omni: Unified Audio Understanding, Generation, and Editing

SIGGRAPH 2026

Audio-Omni is the first end-to-end framework that unifies understanding, generation, and editing across general sound, music, and speech domains.

Capabilities

• Understanding: Audio/video captioning, question answering
• Generation: Text-to-Audio, Text-to-Music, Video-to-Audio, Video-to-Music, Text-to-Speech, Voice Conversion
• Editing: Add, Remove, Extract, Style Transfer

Links

• GitHub Repository
• Project Page
• Model on Hugging Face