Cut Audio JK🐉

Jake's Wan "Cuts" workflows made audio-first video generation a real thing: you define scene cuts, generate a video segment per cut, and stitch them so the visuals land on the beat. Cut Audio JK is the atomic tool that whole family is built on - it takes an AUDIO tensor and slices out the exact window you ask for, by start and end time. It's the audio equivalent of Image Crop, and it's the one node you'd reach for whenever a clip needs to match a specific span of audio.

Unlike the other JK audio cutters, this one doesn't need scene-cut metadata - you type in seconds. Feed it start_time and end_time (defaults 0 and 5, step 0.001 for millisecond precision) plus your audio, and you get back a trimmed audio and the resulting duration as a float, which is handy if you're computing frame counts downstream (multiply duration by fps).

The two flags are where the cleverness lives:

• add_start_mute / add_end_mute - when on, the node pads with silence if your window reaches before 0.0s or past the end of the file, instead of just clamping. That's how these workflows keep every segment the same length even when the source audio runs short - the video sampler gets a full-length audio track to sync against.
• The error you'll actually hit: start_time must be less than end_time. It raises a clear ValueError, not a silent clamp.

The genuinely surprising feature: leave audio unconnected and the node generates a silent clip at 44.1kHz for your requested window. That was added in v2.6.3 and it's the escape hatch for the "no audio available" case - your video-generation loop still gets a correctly-sized silent track instead of crashing. It's also useful for purely visual workflows that want silence as a placeholder.

Install. Part of ComfyUI-JakeUpgrade:

cd ComfyUI/custom_nodes
git clone https://github.com/jakechai/ComfyUI-JakeUpgrade
cd ComfyUI-JakeUpgrade
pip install -r requirements.txt

Or ComfyUI Manager → "JakeUpgrade". Windows portable: install.bat or ../../../python_embeded/python.exe -s -m pip install -r requirements.txt. Restart after. Deps are light (opencv, piexif, simpleeval, PyYAML, toml) - the AUDIO type itself comes from the pack's audio module.

Gotchas: audio in ComfyUI is just {waveform, sample_rate} - if your source was loaded by a Video Helper Suite or similar node, the sample rate carries through, but the silence fallback always uses 44.1kHz. If you later concatenate that silent track with a 48kHz source, you'll have two segments on different clocks - fine in isolation, a silent source of confusion if something downstream assumes one rate. And the family rule: if a JK node reports missing, check config.ini's LOAD_DEPRECATED_NODES flag; the README also says to disable ComfyUI MultiGPU, which misbehaves on current ComfyUI.