ComfyUI-LTXVideo: Advanced LTX-2 Nodes for Video, HDR, and Dubbing in ComfyUI

What ComfyUI-LTXVideo Adds to the Base LTX-2 Integration

LTX-2, the video generation model from Lightricks, is built into ComfyUI core at comfy/ldm/lightricks. Any ComfyUI installation can use LTX-2 for text-to-video and image-to-video generation without installing anything extra. ComfyUI-LTXVideo is a separate custom node package that adds features not in the core integration: IC-LoRA conditioning modes, HDR video output, speech dubbing, a spatial upscaler, and a set of example workflows for LTX-2.3.

The repository describes itself as a collection of custom nodes extending ComfyUI's capabilities specifically for LTX-2's advanced features. It is not a replacement for the core LTX-2 nodes but an extension layer. Users who only need basic text-to-video and image-to-video do not need this package.

The target user is a ComfyUI practitioner working on professional or creative video production: someone who needs depth-guided or edge-guided conditioning, wants to output HDR video for color grading, needs to dub or rephrase speech in a video, or wants to upscale generated clips with generative detail rather than interpolation.

The IC-LoRA System: Four Conditioning Modes

IC-LoRA (Image Conditioning LoRA) is the mechanism for conditioning LTX-2 generation on control signals beyond the text prompt. ComfyUI-LTXVideo provides four IC-LoRA variants.

The Union IC-LoRA combines depth and edge (canny) control into a single LoRA. The README explains that it operates on downsampled latents, which reduces memory usage and speeds up inference while maintaining quality. A single LoRA handles multiple control signals by learning to identify which are present in the input.

The HDR IC-LoRA upgrades SDR video to ACEScct HDR. The README describes it as outputting frames in LogC3-compressed space; the LTXVHDRDecodePostprocess node decodes these back to linear HDR values. The node outputs both a Reinhard-tonemapped SDR preview and the raw linear HDR tensor. EXR export requires OpenImageIO:

pip install openimageio

OpenImageIO is also listed in the repository's requirements.txt.

The Dub-It IC-LoRA generates new lip movements and audio matching a target text prompt while preserving the speaker's identity. It supports both cross-language dubbing and same-language rephrasing. The pipeline is two stages: Stage 1 generates video and audio at base resolution; Stage 2 upscales while freezing the audio.

The Pixel Spatial Upscaler IC-LoRA re-renders low-resolution clips at 2x or 4x by synthesising fine detail. The README calls it a creative upsampler rather than a pixel-accurate refiner: it generates texture and structure from the reference rather than faithfully preserving every pixel.

Installing via Comfy Manager and Example Workflows

The recommended installation path is through Comfy Manager, the standard ComfyUI extension manager:

1. Open ComfyUI. 2. Click the Manager button or press Ctrl+M. 3. Select Install Custom Nodes. 4. Search for LTXVideo. 5. Click Install and wait for it to complete. 6. Restart ComfyUI.

Once installed, the nodes appear in the node menu under the LTXVideo category. Required models are downloaded on first use.

The installation includes example workflows. After installation, they are available at:

ComfyUI/custom_nodes/ComfyUI-LTXVideo/example_workflows/

The LTX-2.3 workflows in the 2.3/ subdirectory cover: text/image to video (single stage, full and distilled model), text/image to video (two stages with upsampling), IC-LoRA for depth/pose/edges, IC-LoRA for motion tracking, IC-LoRA HDR, IC-LoRA Dub-It with two-stage upsampling, IC-LoRA pixel spatial upscaling, and text-to-audio single stage. Older LTX-2.0 workflows cover text-to-video, image-to-video, video-to-video detailing, and the original IC-LoRA conditioning mode.

Hardware Requirements and Dependencies

The README states three prerequisites: ComfyUI installed, a CUDA-compatible GPU with 32GB or more of VRAM, and 100GB or more of free disk space for models and cache.

The 32GB VRAM requirement is the most significant constraint. The Union IC-LoRA specifically reduces this burden by operating on downsampled latents, but the base requirement still applies for full model loading. The repository includes a low_vram_loaders.py file, visible in the top-level directory listing, which may provide paths for lower-VRAM operation; the README's quick-start section does not document it.

Python dependencies declared in requirements.txt include colour-science>=0.4.4 (required for HDR primaries), diffusers, einops, huggingface_hub>=0.25.2, kornia, ninja~=1.11.1.4, openimageio (required for HDR EXR read/write), and transformers[timm]>=4.50.0. The README notes that openimageio and colour-science have no fallback implementations; they are required for the HDR features to function.

Comparison with the Base LTX-2 Core and License Status

Other video-generation models available as ComfyUI custom nodes (such as those from Stability AI) generate video through diffusion without the IC-LoRA conditioning modes that ComfyUI-LTXVideo provides. The HDR output pipeline, the Dub-It speech regeneration, and the Union IC-LoRA depth-plus-edge control are specific to this package and are not replicated by the base ComfyUI LTX-2 integration.

The practical comparison for most users is between this package and the core LTX-2 nodes in ComfyUI. The core nodes handle standard text-to-video and image-to-video workflows. ComfyUI-LTXVideo is needed only for the advanced conditioning, HDR, dubbing, and upscaling capabilities. Teams building production pipelines should start with the core integration and add this package when they need a specific feature it provides.

The license field in the repository metadata is listed as NOASSERTION, indicating that the automated license detection could not identify the license. The repository does include a LICENSE file in the top-level directory, but the README does not quote its terms. Verify the license before using the package in commercial production. The last push was on 2026-09-24, and the repository is under active development.

Editorial conclusion

ComfyUI-LTXVideo is the right package for ComfyUI users who want to go beyond the base LTX-2 text-to-video and image-to-video capabilities: the IC-LoRA modes (depth, edges, motion tracking), the HDR pipeline, and the Dub-It dubbing are not in the core ComfyUI LTX-2 integration. The 32GB+ VRAM requirement is a hard constraint; users with less VRAM should consult the low_vram_loaders.py node mentioned in the repository layout before assuming the package will not run. The license field in the repository is listed as NOASSERTION, meaning the license was not automatically identified; verify the terms with Lightricks before using the package in a commercial workflow.

Frequently asked questions

Official sources

1. Issues
2. Lightricks/ComfyUI-LTXVideo on GitHub
3. Project website
4. README