CoolFace
Modelpublic

Zlikwid/LTX_2.3_Upscale_IC_Lora

sourceHugging Faceupdated 5mo agoView on Hugging Face
73likes
Model Card

LTX 2.3 Upscale IC-LoRA

<video controls autoplay loop muted playsinline width="720"> <source src="https://huggingface.co/Zlikwid/LTX2.3UpscaleICLora/resolve/main/Examples/example1.mp4" type="video/mp4"> </video> <video controls autoplay loop muted playsinline width="720"> <source src="https://huggingface.co/Zlikwid/LTX2.3UpscaleICLora/resolve/main/Examples/example2.mp4" type="video/mp4"> </video> <video controls autoplay loop muted playsinline width="720"> <source src="https://huggingface.co/Zlikwid/LTX2.3UpscaleICLora/resolve/main/Examples/example_3.mp4" type="video/mp4"> </video>

What it does

This LoRA is meant for taking a lower-res or soft video and running it through LTX 2.3 as a second pass to get a cleaner, more detailed result.

It is not a normal upscaler. It is a generative refinement LoRA, so the output can change depending on your workflow/settings.

How to use it

The basic idea is:

text
low-res video → bicubic upscale → run through LTX 2.3 with this LoRA

Before using the LoRA, resize your source video with bicubic scaling to the resolution you want to work at.

Example:

text
640x360 → bicubic upscale to 1280x720 → LTX 2.3 upscale pass

Then use the bicubic-upscaled video as the input/reference video in your LTX 2.3 IC-LoRA workflow.

Use this prompt:

text
upscale

That is the prompt this LoRA is meant to use.

Notes

The main checkpoint is:

text
ltx2.3_upscale_ic-lora_06250.safetensors

Other checkpoints are in:

text
Other Weights/

Those are just there for testing/comparison.

Results will depend on your source video, resolution, denoise/strength settings, frame count, and workflow.