PlatformNSFW

ComfyUI LipDub workflow syncs video to audio in new tutorial

ComfyOrg released a recorded tutorial demonstrating LipDub, a ComfyUI workflow that aligns generated video output with audio tracks using phoneme extraction and frame-by-frame conditioning.

May 19, 2026

ComfyUI LipDub workflow syncs video to audio in new tutorial

ComfyOrg released a recorded tutorial demonstrating LipDub, a ComfyUI workflow that syncs generated video output with audio tracks. The stream walks practitioners through setup, node configuration, and real-time debugging for lip-sync animation tasks—a capability previously available only through closed platforms like Runway and Synthesia.

More in Platform

Watch: ComfyUI LipDub workflow syncs video to audio in new tutorial

LipDub runs inside ComfyUI's node-based interface, letting users wire audio input directly into video generation pipelines. The workflow extracts phonemes from an audio file, converts timing data into per-frame conditioning signals, and feeds those into any open-weight video model that accepts frame scheduling. The tutorial demonstrates the same workflow running on both AnimateDiff and LTX checkpoints, showing how the node graph adapts to different model architectures without modification.

The workflow is available as a JSON export on ComfyOrg's GitHub repository. Local ComfyUI users can import the file and swap in their own audio and model checkpoints without editing the core node graph. The recording includes timestamps for each setup step and troubleshooting segment, addressing common pitfalls around timing drift and audio preprocessing.

Local video synthesis without closed APIs

ComfyUI's extensibility through custom nodes has made it the de facto standard for practitioners running open-weight models like AnimateDiff, LTX, and Hunyuan Video on consumer hardware. LipDub extends that ecosystem into audio-driven animation, removing what had been the last major barrier to producing character animation with full creative control on local hardware. The workflow runs entirely on user hardware—no external API calls, no licensing fees—making it accessible to anyone with a capable GPU and the patience to configure the node graph.