IAMCCS Audio Phase Lazy Gate

If you've ever built one of those giant LTX-2 audio workflows, you know the problem this node exists for: the graph is one long chain, and running one stage of it - say, just the dialogue-to-audioboard export - still forces ComfyUI to execute the whole upstream chain first, because everything's hard-wired. IAMCCS_AudioPhaseLazyGate is a lazy gate that splits a big audio pipeline into named phases, so you can run just the part you're iterating on.

The idea is simple. You drop a gate into the chain and set phase to one of three stages: dialogue_to_audioboard, publish_to_shotboard, or video_from_shotboard. Those map to the IAMCCS production path from the pack README (Shotboard/AudioBoard workflow): first generate and assemble the dialogue audio, then publish it into the Shotboard timeline, then generate video from that shotboard. The gate passes a cine_linx payload and an audio_timeline_json string through - those are IAMCCS's custom "linx" supernode connectors that carry the whole audio timeline, tracks, and metadata in one object - and stamps which phase you're in onto the payload so downstream nodes can react.

The "lazy" part is what makes it useful. The gate's inputs are marked lazy, and when enabled is off it tells ComfyUI not to execute its upstream inputs at all (the source calls this "did not request upstream lazy inputs"). So you can keep a big graph fully wired, flip a gate's enabled switch, and ComfyUI only runs the branch you care about. Each node also outputs an enabled copy and a report JSON string describing what it decided.

What you actually set, as a beginner:

• phase - which stage of the audio pipeline this gate guards. Pick the one you're running.
• enabled - the master switch (labeled ACTIVE/BYPASS in the UI). Off = lazy passthrough with no upstream work.

Outputs you'll wire: cine_linx and audio_timeline_json flow on to the next audio or Shotboard node; report is your debugging friend when a stage runs but something downstream looks empty.

Installing it: this is one class in IAMCCS/IAMCCS-nodes, so it installs with the whole pack. ComfyUI Manager → search "IAMCCS", or cd ComfyUI/custom_nodes && git clone https://github.com/IAMCCS/IAMCCS-nodes.git, then restart ComfyUI. No models or extra pip packages to grab - it's plain torch under the hood, and the pack's only real demands are a modern ComfyUI (≥ 0.3.0) with Python 3.12+ and Torch 2.8+.

Where people get burned: the phase names are the author's own pipeline vocabulary, and if your workflow isn't following the Shotboard/AudioBoard structure the three choices won't map cleanly. Treat the gate as a routing label, not a magic stage splitter - it doesn't know what's upstream, it just controls whether upstream executes and stamps the phase name onto the payload. That's a feature (it's fast and predictable), just don't expect it to infer your graph. If a stage runs but the output audio is empty, check the report on the gate first - it tells you whether the lazy request was honored.

One more note: the author (Carmine "IAMCCS" Cristallo Scalzi) builds these as a coherent system and documents them via the pack README plus Patreon deep-dives, and the community has real respect for the pack's video work - IAMCCS-nodes' SVI Pro v2 support got a 200+ upvote thank-you thread on r/comfyui. The gate is a small piece of that system, so it shines when you use it the way the linked Shotboard workflows do.