Add Realistic Audio to AI Video! ComfyUI MMAudio Workflow

This ComfyUI MMAudio workflow is for adding realistic audio to AI video; it takes a video path, runs MMAudio loading and sampling stages, previews audio, and includes a video-combine output stage.

Seven node types cover video loading and information, MMAudio feature and model loading, audio sampling and preview, and video combining.

Video handling uses VHS_LoadVideoPath and VHS_VideoInfo; output-side nodes include PreviewAudio and VHS_VideoCombine.

MMAudioFeatureUtilsLoader, MMAudioModelLoader, and MMAudioSampler provide the feature, model-loading, and sampling stages.

The workflow is published on Civitai by Lordcaocao2025.

The workflow lists four model files: apple_DFN5B-CLIP-ViT-H-14-384_fp16.safetensors, mmaudio_large_44k_v2_fp16.safetensors, mmaudio_synchformer_fp16.safetensors, and mmaudio_vae_44k_fp16.safetensors.

It also lists two node packs: comfyui-mmaudio and comfyui-videohelpersuite.

Before use, confirm that all four listed model files are available under the exact filenames shown.

Suggestion · not verified

The Civitai listing contains an embedded YouTube video but no distinct written description, so verify setup details before use.

Suggestion · not verified