ReLTXVPatcherAdvanced

The block-selectable version of ReLTXVPatcher. Same purpose - patching an LTX Video model so RES4LYF's style-transfer and dual-conditioning system can operate on it - but instead of an all-or-nothing toggle, you choose which of the model's transformer blocks the style guide is allowed to reach.

Why you'd reach for this over the plain version

For video-to-video style transfer specifically, letting a style guide touch every block of the network can pull temporal consistency around in ways you don't want - a style signal that's too aggressive on later blocks can flicker frame to frame even when it looks fine on a single still. Narrowing which blocks get patched is the lever for dialing that back: keep the guide in earlier double-stream blocks for a more compositional, stable influence, and be more cautious pushing it into later single-stream blocks if you're seeing per-frame instability. This is genuinely trial-and-error - there's no published map of which LTXV block ranges correspond to which effect, and that map would differ across the 1.x/LTX-2/LTX-2.3 generations anyway.

The inputs and outputs that matter

• model (MODEL, required) - your loaded LTX Video model.
• doublestream_blocks (STRING, default "all") - comma-separated indices or ranges of double-stream blocks to patch.
• singlestream_blocks (STRING, default "all") - same, for single-stream blocks.
• style_dtype (default float64) - precision for the style computation, separate from your loaded model's own dtype.
• enable (BOOLEAN, default true) - bypass switch.

Output: a patched model.

How to install it

Ships 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 need a working LTXV setup already in place - your own model weights and, in most cases, Lightricks' ComfyUI-LTXVideo node pack for the surrounding pipeline. This node patches into an existing LTXV workflow rather than building one.

Common issues & troubleshooting

Style transfer is flickering across frames. This is the case the advanced node exists for: try narrowing singlestream_blocks away from "all" first, since later blocks are the more likely source of per-frame instability, and re-test on a short clip before committing to a full render.

Block strings not doing anything. No validation on these free-text fields - a bad range or a typo fails quietly. Confirm with the plain patcher (or "all" on both fields here) that the style effect is present at all before troubleshooting a specific narrowed range.

Confusing errors that turn out to be an LTXV version mismatch. Check this before anything else - 1.x, LTX-2, and LTX-2.3 use different node sets and (as of 2.3) an incompatible VAE from the previous generation. Make sure model, LTXV nodes, and text encoder all match generations.

You're not sure narrowing blocks is worth the effort. If the plain ReLTXVPatcher already gives you the result you want, stop there - this node is for the specific case where the default style influence is too strong on temporal consistency or fine detail and you need a scalpel instead of a switch.