Latent to RawX

This one isn't a beginner node, and it's worth saying so up front rather than pretending otherwise. Latent to RawX pulls out raw_x - the internal, un-decoded diffusion-space tensor that RES4LYF's samplers carry alongside the "normal" latent - and hands it back to you as its own LATENT. If you don't already know why a sampler would carry hidden state on top of its visible output, you probably don't need this node yet, and that's fine.

Here's the context that makes it make sense. A regular ComfyUI LATENT is just a tensor. But RES4LYF's own samplers support chaining - running, say, the first half of your steps in one sampler node and the rest in a second one, a real and fairly common pattern for things like Wan 2.2's split high-noise/low-noise passes. For that to work cleanly, the latent passed between the two stages needs to carry more than the visible image data: it needs the actual unprocessed diffusion state (raw_x) the next stage should resume from, not a decoded approximation of it. This node lets you reach in and extract that hidden state directly.

How it works

You feed it a latent that already carries this extra state - typically one that came out of a RES4LYF sampler node rather than an Empty Latent Image - and it returns just the raw_x component as a standalone latent, discarding everything else attached to it.

The inputs and outputs that matter

• latent (required, and the only input) - a latent carrying RES4LYF's internal sampler state.
• Output: latent_raw_x - the extracted raw diffusion-space tensor, as a LATENT you can inspect, pass along, or feed into another node in the same family (Latent-Replace-State-Info, Latent-Transfer-State-Info).

How to install it

Via ComfyUI Manager: search "RES4LYF" and install. Manually, from ComfyUI/custom_nodes:

git clone https://github.com/ClownsharkBatwing/RES4LYF/
cd RES4LYF
pip install -r requirements.txt

Activate your venv first if you use one; on a portable install, use python_embedded\Scripts\pip.exe in place of pip. Restart ComfyUI afterward.

Common issues & troubleshooting

Honest note before anything else: this is genuinely obscure plumbing. There's no written explanation of it beyond the node's name and its single input/output, and no community threads discuss raw_x by name - it doesn't show up in general Stable Diffusion discussion at all. You're far more likely to encounter this node already wired into someone else's advanced chainsampler or video workflow than to add it to a blank canvas yourself.

The output looks like noise, not an image. That's expected - raw_x is the raw diffusion-space state, not something meant to be decoded and viewed directly. If you want to preview it, decode it through a VAE the same way you would any other latent, but don't expect it to look like a normal generation partway through.

You fed it a latent from Empty Latent Image or a plain image encode. Those don't carry RES4LYF's internal state at all, so there's nothing meaningful for this node to extract. It's meant for latents that came out of one of the pack's own sampler nodes.

You're trying to understand this from the node alone. Don't - open one of the example workflows linked from the RES4LYF README (the "intro to clownsampling" workflow is the best starting point) and see how the pack's own chained-sampler setups use state-info nodes like this one in context.