LC Pipe → AV Pipe 🔌

If you've used lonecatone23's image pack (ComfyUI_LC123_nodes), you know the LC_PIPE: one wire that carries your prompts, seed, steps, sampler, and size down the graph. This AV pack speaks a different dialect, LC_AV_PIPE, and LC Pipe → AV Pipe is the converter between them - you feed it an LC_PIPE and it hands back an LC_AV_PIPE carrying everything the two have in common.

The distinction exists for a good reason. The image pack's pipe can hold models, CLIP, VAE, latents and images; this audio/video pipe is a leaner animal built around video frames, audio, fps and duration. The author's rule is explicit: never put video into an LC_PIPE, and the two pipes simply don't share those heavy slots. So instead of trying to make one pipe do everything, this node copies only the shared fields - prompts, seed, steps, sampler, and size - into a fresh AV pipe, and leaves everything video-specific out. If you were imagining it would teleport your model and CLIP over, it won't, and that's deliberate; the tooltip lists exactly what does not transfer.

How it works

Mechanically it's a clean dict hand-off: it reads the incoming LC_PIPE, copies the common keys (width, height, positive_prompt, negative_prompt, seed, total_steps, sampler_name, scheduler, denoise, cfg…), normalizes aliases (the AV pipe wants total_steps/sampler_name, which the image pipe may store as steps/sampler), and never mutates the source pipe. One input socket, one output socket, zero widgets - this is as frictionless a bridge as the pack ships.

Where you'd use it

The wiring pattern it enables: generate your still with the image pack's LC_PIPE carrying all the generation metadata, then convert to LC_AV_PIPE so an LC Create Video, LC Save Video, or LC Save Video Metadata downstream sees the same seed and prompt without you re-entering them. It's also how you keep the metadata honest when your video started as an image generation - the seed that made the first frame rides along into the video's save metadata.

How to install it

Ships in lonecatone23/ComfyUI_LC_AV_nodes (LC Audio_Video nodes, MIT - companion to ComfyUI_LC123_nodes, installed as a separate repo):

• ComfyUI Manager: search "LC Audio_Video" or ComfyUI_LC_AV_nodes.
• Manual: cd ComfyUI/custom_nodes && git clone https://github.com/lonecatone23/ComfyUI_LC_AV_nodes, restart ComfyUI (console: [LC AV] total 16 nodes).

No extra dependencies or downloads.

Where it bites

The trap is expecting a lossless conversion. Because only shared keys transfer, anything unique to LC_PIPE (models, CLIP, VAE, latent, image, mask) is simply not there when you unpack the AV pipe later - so don't wire the AV side into a node that needs a model and expect the converter to have smuggled one across. And note this node takes an LC_PIPE, not an LC_AV_PIPE; if you already have an AV pipe and just want to enrich it with an image pipe, that's what LC AV Pipe (in/edit) is for.