AD_LTX_config

This node does one job and does it well: it takes the single config bundle that AD_LTX_load_model builds and breaks it back out into every individual piece - model, clip, the image latent/VAE pair, the audio latent/VAE pair, and the raw width/height/frame_rate/frames_number values as plain numbers. It's the "bus splitter" half of the Loader-Controller-Sampler design the pack's README talks about: the loader packs everything into one wire so your graph doesn't turn into spaghetti, and this node unpacks it again wherever you need the individual pieces - say, feeding width/height into a resize node, or grabbing just model to pass through a LoRA stack before it hits the sampler.

How it works

There's no processing happening here - no sampling, no VAE work, nothing that costs GPU time. It's a pure pass-through/unpacker. Given that, it's a cheap node to insert anywhere in a graph that you need one specific field out of a config bundle without breaking the rest of the chain.

The inputs and outputs that matter

• config (LTX_CONFIG) - the one required input, coming from AD_LTX_load_model (or anything else in the pack that emits the same type).

Outputs, all unpacked from that one input: config (re-emitted, so you can keep chaining), model (MODEL), clip (CLIP), img_latent (LATENT), img_vae (VAE), audio_latent (LATENT), audio_vae (VAE), width (INT), height (INT), frame_rate (INT), frames_number (INT).

How to install it

Through ComfyUI Manager: search "ComfyUI-Apt_Preset". Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/cardenluo/ComfyUI-Apt_Preset.git

Then install.bat (Windows, double-click) or pip install -r requirements.txt (Linux/Mac), and restart ComfyUI. This node needs no model files of its own - everything it outputs came from whatever's plugged into config, so its install requirements are entirely inherited from AD_LTX_load_model upstream.

Common issues & troubleshooting

Nothing plugged into config. This node has exactly one required input and it's not optional - without an upstream AD_LTX_load_model (or equivalent) feeding it a real LTX_CONFIG object, there's nothing for it to unpack.

Not sure whether you need this node at all. If you're only ever going to use the bundled outputs together (say, feeding model/img_latent/img_vae straight into AD_LTX_sampler as a group), you likely don't need this - AD_LTX_sampler and friends can often take the bundle directly or via their own required inputs. Reach for AD_LTX_config specifically when you need to peel off one field, like grabbing width/height as plain integers to drive a math node or a resize step elsewhere in the graph.

Values look stale after changing something upstream. Since this is a pure unpacker with no caching logic of its own, if outputs seem out of date, the more likely culprit is ComfyUI's own execution caching on the upstream AD_LTX_load_model node not re-running - force a re-run there rather than assuming this node is holding onto old data.