ReLTXVPatcher

LTX Video is the speed model of the open video world - Lightricks built it to generate real-time-ish, and people run it specifically because it's fast enough to iterate with on modest hardware, not because it's the highest-fidelity option (that's usually Wan). RES4LYF's video-2-video and style-transfer support extends to LTXV, and ReLTXVPatcher is what hooks the pack's style-guide machinery into an LTX model so it plays correctly with that pipeline.

What it does

RES4LYF's style-transfer and dual-conditioning nodes need to inject an extra guidance signal into the model's forward pass, and video diffusion transformers have their own internal block structure that a generic patch can't handle correctly. This node wraps a loaded LTXV model so RES4LYF's style/guide system knows how to talk to it. It's the plain on/off version; if you need to control exactly which transformer blocks the style signal reaches, that's ReLTXVPatcherAdvanced.

Worth knowing going in: LTXV comes in a genuinely fragmented lineup - 1.x (2B/13B, dev or distilled), LTX-2 (19B), and LTX-2.3 (22B) - and nodes, weights, and even the VAE differ across those generations. This patcher is built for whichever LTXV weights and node ecosystem you actually have installed; if something errors oddly, version mismatch inside the LTXV family itself is worth ruling out before blaming RES4LYF.

The inputs and outputs that matter

• model (MODEL, required) - your loaded LTX Video model.
• style_dtype (default float64) - precision for the internal style computation, separate from whatever dtype your LTXV weights are running at.
• enable (BOOLEAN, default true) - bypass switch, so you can compare with and without the patch active without rewiring.

Output: a patched model, wired onward to your video sampler like any model output.

How to install it

Comes with RES4LYF - no separate install. Via ComfyUI Manager: search "RES4LYF", install, restart. Manually:

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

You'll separately need LTXV's own weights and, for full functionality, Lightricks' own ComfyUI-LTXVideo nodes - this patcher assumes you already have a working LTXV generation pipeline and slots into it, it doesn't set one up for you.

Common issues & troubleshooting

Nothing changes when you toggle it. Confirm enable is on, and that your workflow actually routes through RES4LYF's style-guide or dual-conditioning nodes downstream - this patch is inert on its own if nothing consumes the style signal it enables.

Version confusion inside LTXV itself. This is the single biggest source of "why doesn't this work" in the LTXV world generally, patcher or not: 1.x, LTX-2, and LTX-2.3 use different node sets and incompatible latent spaces (2.3's VAE was rebuilt from LTX-2's). Make sure your model, your LTXV nodes, and your text encoder are all from the same generation before troubleshooting anything else.

Slower than you expected. LTXV's whole reputation is built on being the fast option, so if generation feels sluggish, check the obvious things first - whether you're on a distilled variant, whether Q8 acceleration is actually available on your GPU (restricted to RTX 40-series), and whether offloading flags are fighting your VRAM headroom - before assuming this patcher's float64 style computation is the bottleneck. It's a small addition next to the video model itself.