ComfyUI-LTXVideo: Lightricks’ Extra Nodes for LTX-2 Video

LTX-2 already runs inside ComfyUI’s core. ComfyUI-LTXVideo is where Lightricks puts everything else: ready-made workflows, control LoRAs and a handful of specialised nodes for things like HDR output and dubbing. If you generate video in ComfyUI and use LTX models, this is the pack the model’s own authors maintain.

What it does

The repository is a set of custom nodes and example workflows for the LTX-2 family of video models. The README lists workflows for LTX-2.3 covering text- or image-to-video in one or two stages (the two-stage versions add upsampling), plus older LTX-2.0 workflows for text-to-video, image-to-video and a video-to-video detailer.

Beyond the basics, the pack documents several IC-LoRA add-ons:

• Union control — one LoRA that accepts depth or edge (canny) maps, working on downsampled latents to cut memory use.
• HDR — generates linear HDR video encoded in ARRI LogC3, with an optional 16/32-bit EXR sequence export for grading.
• Dub-It — takes a source video and new dialogue text, then regenerates lip movement and audio, either translating or rephrasing while keeping the speaker’s voice.
• Pixel spatial upscaler — 2× and 4× variants that re-render low-resolution clips with generated detail. The README is explicit that this is a creative upsampler, not a pixel-accurate one.

Who it’s for

• ComfyUI users who already work with LTX video models and want the maintainer’s reference workflows.
• Post-production people interested in HDR/EXR output or AI dubbing inside a node graph.

Getting started

According to the README, you need ComfyUI installed, a CUDA-compatible GPU with 32GB or more of VRAM, and at least 100GB of free disk space for models and cache. Installation goes through ComfyUI Manager: open the Manager, choose Install Custom Nodes, search for “LTXVideo”, install and restart. The nodes then appear under an “LTXVideo” category, and the required models download on first use. Example workflows are installed to ComfyUI/custom_nodes/ComfyUI-LTXVideo/example_workflows/.

How to install ComfyUI-LTXVideo for LTX-2.3 (October 2026)

First check whether you need custom nodes. ComfyUI supports native LTX-2.3 text-to-video and image-to-video templates. For a basic generation, update ComfyUI, open the workflow template browser, search for LTX-2.3, select a template and follow its model-download prompts. Install the separate Lightricks ComfyUI-LTXVideo node pack when you specifically need its additional nodes or maintained example workflows, such as IC-LoRA controls, HDR, Dub-It or two-stage upscaling. ComfyUI’s LTX-2.3 announcement documents native support.

Method A: install or update with ComfyUI Manager

• Update ComfyUI first, and back up your working workflows and custom-node versions.
• Open ComfyUI Manager, search for ComfyUI-LTXVideo, and verify the publisher repository is Lightricks/ComfyUI-LTXVideo.
• Choose Install or Update, restart ComfyUI, and reload the LTX-2.3 example workflow.
• Follow the workflow’s missing-model prompts. Required checkpoints, text encoders, LoRAs and upscalers vary by workflow; do not assume one download list fits every graph.

Method B: manual custom-node installation

For a self-managed ComfyUI installation, open a terminal in ComfyUI/custom_nodes/ and run:

git clone https://github.com/Lightricks/ComfyUI-LTXVideo.git
cd ComfyUI-LTXVideo
python -m pip install -r requirements.txt

Run the pip command using ComfyUI’s own Python environment, not an unrelated system Python. Windows Portable users must use its bundled interpreter. To update an existing Git clone, run git pull in that custom-node folder, then reinstall requirements with the same ComfyUI interpreter and restart. See the official LTX ComfyUI integration guide.

Which LTX-2.3 workflow should you load?

In ComfyUI/custom_nodes/ComfyUI-LTXVideo/example_workflows/2.3/, the official repository provides one-stage and two-stage text/image-to-video workflows, distilled-model workflows, and specialist IC-LoRA workflows for pose, depth, edges, motion tracking, HDR, dubbing and spatial upscaling. Start with the simpler text-to-video or image-to-video example; use two-stage upscaling only after a basic test succeeds. The project README lists a CUDA-compatible GPU with 32GB+ VRAM and 100GB+ free disk space as its recommended starting requirements, not a universal minimum for every native ComfyUI variant.

Troubleshooting: missing nodes, models and CUDA errors

• LTXVideo nodes missing: verify the package installed into the active ComfyUI instance, inspect startup logs and restart after installation.
• Model not found: use the exact checkpoint and encoder requested by the loaded workflow; check the model selector and folder paths before downloading unrelated checkpoints.
• CUDA out of memory: reduce frame count or resolution, try an appropriate distilled/quantized workflow where documented, or switch to a cloud GPU. Do not assume that any model variant runs on a low-VRAM card.
• Broken workflow after update: compare your installed node version with the example workflow version and consult the repository issues; avoid upgrading custom nodes blindly in a production project.

For the broader model family, read our LTX Video overview. For the underlying graph and video export stages, see ComfyUI Video Workflows. These instructions are documentation-based; The Signal has not yet measured performance on a test GPU.

License status

Limitations and open questions

• The 32GB+ VRAM requirement puts it out of reach of most consumer GPUs.
• 119 open issues on GitHub as of 15 September 2026.
• We have not run these workflows; quality and speed claims in this profile are the maintainer’s descriptions, not our measurements.

Sources

• GitHub repository and README
• License file
• Main LTX-2 repository (linked from the README)

More open video tooling: ComfyUI · Video Generation.

This profile was drafted with AI assistance from the project’s public repository and reviewed by The Signal’s editors before publication. See our AI Disclosure.