MschA2V Load Audio Path
Load your song as audio you can actually sync to
- audio
- duration_seconds
Every audio-driven pipeline has to start somewhere, and in the MSCH A2V suite that somewhere is MschA2V Load Audio Path. It's the most boring node in this article set on purpose: you type a path to a song file, it returns ComfyUI's native AUDIO plus the song's duration in seconds, and optionally applies gain. Boring is the point - it's the node that gets your soundtrack into the graph as something the beat detectors, planners, and samplers can all read from one socket.
Where it sits
MSCH A2V is the music-video side of the MSCH Nodes pack: an orchestration layer that turns a song plus beat-snapped prompts into a video generated by MiniMax H3 - the open-weights 33B model that generates video with native audio rather than bolting a silent clip to a soundtrack. The A2V chain is roughly: load the song, sequence prompts to the beats, plan the shots, sample through H3, assemble. LoadAudioPath is step one. A second entry point matters here: you can connect audio directly to other A2V nodes, but the extra duration_seconds output is exactly what a beat sequencer needs to bound its timeline, and this node exists to hand both out at once.
Inputs and outputs
The only required input is audio_path - a full local filesystem path (the file has to exist on the machine running ComfyUI; this isn't a URL loader). gain_db (optional, −60 to +24 dB, default 0) trims or boosts the level before it reaches the rest of the graph - useful when your song is quiet in a mix and you want the analysis to see it clearly, though don't push past a few dB if you care about the output staying clean.
Outputs: audio (AUDIO) feeds the sampler chain and, later, the encoder for muxing; duration_seconds (FLOAT) feeds the prompt sequencer or anywhere else that needs the song length.
Install and prerequisites - read this before you blame the node
The A2V component only loads its full behavior when you've also installed the separate comfyui-minimax-h3-audio-T8 integration and have compatible H3 model/VAE checkpoints loadable in your ComfyUI - MSCH A2V is an orchestration layer on top of that H3 port, not a replacement for it. The pack itself is the usual unified install:
cd ComfyUI/custom_nodes
git clone https://github.com/mariobilly/msch-comfyui-nodes.git
cd msch-comfyui-nodes
python -m pip install -r requirements.txt
or search "MSCH Nodes" / msch-comfyui-nodes in ComfyUI Manager, then restart.
One more honest caveat before you sink hours in: MiniMax H3's community licence excludes the US, EU, UK and South Korea from the applicable territory for the local weights - if you're in one of those regions you're not licensed to run H3 locally, API aside. That's an H3 licence matter, not an MSCH one, but it's exactly the kind of thing you want to discover before building the workflow.
Gotchas
The file-path input trips people up twice: relative paths resolve relative to wherever ComfyUI started, so use absolute paths, and the node doesn't watch the file for changes (a STRING widget, not a file picker, so double-check typos - a missing file fails at execution). Also keep gain_db modest; the beat analysis downstream is more robust to a slightly quiet source than to a clipped one.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| audio_path | STRING | — | |
| gain_dbopt | FLOAT | 0.0-60–24 | — |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| audio | AUDIO | — |
| duration_seconds | FLOAT | — |