Sony Woosh - Longer videos — v1.0

Use Sony Woosh - Longer videos in ComfyUI to generate sound effects, ambient audio, or scene-based audio from a video and create a rendered video with generated audio synchronized to it.

Sony Woosh supports video inputs up to 8 seconds. Longer videos are split into segments, processed individually, then joined into audio synchronized with the original video.

The workflow includes Qwen3.5 vision analysis for prompt generation, frame views for segment timing, audio previews, source-audio mixing, and optional dialogue suppression with SAM3 masking.

Input a video. Automatic Mode analyzes it and generates prompts for each segment; Manual Mode lets you set custom prompts, segment lengths, seeds, and audio behavior.

The output is a rendered video with generated audio synchronized to the original; depending on the selected mode, the original audio can also be mixed in.

When installing the Woosh custom node through ComfyUI Manager, the instructions say to verify that it points to https://github.com/Saganaki22/ComfyUI-Woosh.

The listed node packs include cg-use-everywhere, comfyui_essentials, comfyui-custom-scripts, comfyui-easy-use, and comfyui-impact-pack.

Other listed node packs are comfyui-kjnodes, comfyui-melbandroformer, comfyui-videohelpersuite, ComfyUI-Woosh, and rgthree-comfy.

The workflow also names MelBandRoformer_fp16.safetensors, qwen3.5_9b_bf16.safetensors, and sam3.1_multiplex_fp16.safetensors. Download locations for the Qwen and SAM3 files are not provided, so confirm where to obtain them before setup.

Vorschlag

Choose Automatic Mode for video analysis and prompts generated for each segment, or Manual Mode to set prompts and segment options yourself.

Vorschlag · nicht geprüft

Frame views show segment timing, and audio previews let you review segments before final rendering. Yellow nodes contain configurable options; red nodes contain instructions and notes.

If ComfyUI Manager reports a security-level error while installing Woosh, the instructions suggest temporarily setting security_level = weak, then restoring the original setting and restarting ComfyUI.

The source listing does not provide license information.