LTX-2.5-22b-IC-LoRA-SDR-To-HDR Video Generation Model: File List

October 1, 2026

At a Glance

Values determined by this site’s code when the information was collected. Dates are JST.

Overview

Lightricks has released “LTX-2.5-22b-IC-LoRA-SDR-To-HDR", an adapter for the video generation model “LTX-2.5" that converts SDR (Standard Dynamic Range) videos into HDR (High Dynamic Range) videos. It is an IC-LoRA (in-context LoRA) adapter specialized for the video-to-video task of taking an SDR video as input and outputting an HDR-processed video of the same content. The underlying technology is based on the accompanying paper, “HDR Video Generation via Latent Alignment with Logarithmic Encoding", and is reported to be a method that achieves HDR generation by leveraging the visual priors of a pre-trained video generation model.

Specifications

• Base Model: Provided as an IC-LoRA adapter for Lightricks/LTX-2.5 (a 22B parameter video generation model)
• Task: video-to-video (SDR video to HDR video conversion)
• Distribution Format: safetensors
• License: ltx-2.x-community-license (registered on Hugging Face as license:other). No specific terms of use are stated in this material
• Obtaining access requires agreement on Hugging Face (gated model)
• Training Method (described in the paper): It is reported that a lightweight fine-tuning approach was adopted, using logarithmic encoding widely used in filmmaking pipelines to naturally align the HDR image distribution with the latent space of the generation model without retraining the encoder. Additionally, a strategy incorporating camera-mimicking degradations into training is said to be used, enabling the model to infer high dynamic range information—which cannot be directly observed from the input SDR video—from its learned priors.

Performance and Quality

Benchmark tables and numerical evaluations are not listed on the model card itself. However, the abstract of the supporting paper states that high-quality HDR video generation is possible with minimal adaptation to the pre-trained video model, “achieving strong results diverse scenes and challenging lighting conditions". It is also reported that even in image formation regimes fundamentally different from the training data distribution of standard generative models, such as HDR, treating the representation method to align with the prior distribution allows it to be handled effectively without redesigning the generative model itself. Note that these are descriptions by the authors in the paper, and third-party comparative evaluations or specific numerical metrics are not included in this material.

Strengths and Use Cases

This model specializes in upgrading existing SDR (Standard Dynamic Range) videos into high-quality HDR (High Dynamic Range) videos. Primary intended use cases include remastering legacy video assets and expanding the dynamic range of SDR videos generated by AI.

It is reported to be highly effective in the following situations:

• Restoration of Information-Deficient Scenes: According to the paper, by incorporating camera-mimicking degradations into the training process, the model can infer and complement luminance information lost to clipping or crushing in SDR videos using its visual priors. This reportedly allows the generation of high dynamic range content that is unobservable at the input stage.
• Generation in Complex Lighting Environments: It is stated that HDR conversion can be performed while maintaining robust generative capabilities across diverse scenes and under difficult lighting conditions. This is because it adopts logarithmic encoding, which is standard in filmmaking pipelines, to naturally align the HDR data distribution into the latent space of the generation model.
• Leveraging the Expressiveness of the Base Model: The original model, “LTX-2.5", is a video generation model with a large scale of 22B (22 billion) parameters. This model supports multiple languages including Japanese, and possesses multi-functional foundations capable of generating video from images and text, as well as audio generation. Since this adapter directly utilizes the knowledge of this robust foundation, high-quality conversion while maintaining video consistency can be expected.

Distributed Files

Weight files published in Lightricks/LTX-2.5-22b-IC-LoRA-SDR-To-HDR, listed by this site from the Hugging Face API. Sizes are the actual file sizes.

How to Get It

This model is available on the Hugging Face repository “Lightricks/LTX-2.5-22b-IC-LoRA-SDR-To-HDR".

• Required Procedures for Acquisition: This model is published as a gated model. Obtaining and downloading it requires agreeing to the “ltx-2.x-community-license" on Hugging Face and submitting an access request to the developers.
• Distribution Format: Weight files are provided in safetensors format.
• Operating Environment: Operates as an IC-LoRA adapter for the base model “Lightricks/LTX-2.5". The ltx library is supported.

Can You Run It Locally?

The publisher distributes this model as safetensors.

License — ltx-2.x-community-license (Commercial use allowed with conditions): Lightricks’ LTX-2.x Community License. Businesses with annual revenue of $10 million or more need a paid license for commercial use (Section 2.1); smaller users may use it free of charge.

Access: You must accept the license on Hugging Face before downloading.

Compiled by this site’s code from the published formats and the license field. License summaries are not legal advice — check the publisher’s original terms before relying on them.

Related Articles

• LTX-2.5-22b-IC-LoRA-Restore Video Generation Model: File List
• Lightricks Releases Multiple LoRA and IC-LoRA Adapters for LTX-2.5
• H3-to-LTX-Latent-Adapter: 4GB+ VRAM, File List

What to Read Next

• Formats this model is available in → Safetensors format guide and models
• Learn about the publisher → Lightricks (LTX): models, licenses and articles
• Other models for the same task → Other video generation models

Sources

• https://huggingface.co/Lightricks/LTX-2.5-22b-IC-LoRA-SDR-To-HDR
• https://arxiv.org/abs/2604.11788