CivArchive

    Lightricks LTXV

    Lightricks is now available to Civitai users in both text-to-video and image-to-video modes, within the Civitai Generator! For more details, check out our guide to video in the Civitai Generator!

    Update: 5/8/2025 - New Version! 13b 0.9.7 + 13b 0.9.7 fp8! New sample images, guide, and workflow to follow!

    Update: 4/18/2025 - New Version! 0.9.6 + 0.9.6 Distilled! You can plug 0.9.6 into the same workflows and it'll work, but the Distilled version requires a new workflow/node.

    Update: 3/6/2025 - New Version! 0.9.5! You'll need new workflows to use the new model, examples available here - https://education.civitai.com/civitais-quickstart-guide-to-lightricks-ltxv/#workflows-model-version-0-9-5-latest-model

    Originally posted on Huggingface.

    Read our Lightricks LTXV Quickstart Guide on the Education Hub!

    This model card focuses on the model associated with the LTX-Video model, codebase available here.

    LTX-Video is the first DiT-based video generation model capable of generating high-quality videos in real-time. It produces 24 FPS videos at a 768x512 resolution faster than they can be watched. Trained on a large-scale dataset of diverse videos, the model generates high-resolution videos with realistic and varied content. We provide a model for both text-to-video as well as image+text-to-video usecases.

    General tips:

    • The model works on resolutions that are divisible by 32 and number of frames that are divisible by 8 + 1 (e.g. 257). In case the resolution or number of frames are not divisible by 32 or 8 + 1, the input will be padded with -1 and then cropped to the desired resolution and number of frames.

    • The model works best on resolutions under 720 x 1280 and number of frames below 257.

    • Prompts should be in English. The more elaborate the better. Good prompt looks like The turquoise waves crash against the dark, jagged rocks of the shore, sending white foam spraying into the air. The scene is dominated by the stark contrast between the bright blue water and the dark, almost black rocks. The water is a clear, turquoise color, and the waves are capped with white foam. The rocks are dark and jagged, and they are covered in patches of green moss. The shore is lined with lush green vegetation, including trees and bushes. In the background, there are rolling hills covered in dense forest. The sky is cloudy, and the light is dim.

    Description

    1. Improved model - removes "strobing texture" artifacts and generates better motion.

    2. STG support

    3. Integrated image degradation system for improved motion generation.

    4. Additional initial latent optional input to chain latents for high res generation.

    5. Image captioning in image to video.

    FAQ

    Comments (19)

    AiMetatronDec 19, 2024· 3 reactions
    CivitAI

    This model is god speed!

    jpXerxesDec 20, 2024· 3 reactions
    CivitAI

    is 0.9.1 missing the vae? throws errors in the example workflow, while 0.9 does work

    theallyDec 20, 2024· 3 reactions

    Yes! Annoyingly, Core-ComfyUI does not support this new model - but it's coming soon. If you'd like to run the new model before then you need the official LTXV nodes and workflows.

    jpXerxesDec 20, 2024· 1 reaction

    @theally thx

    LG89Dec 20, 2024· 1 reaction
    theallyDec 20, 2024· 2 reactions

    @LG89 Awesome! Thanks for letting me know!

    CavernustDec 26, 2024
    CivitAI

    I have a problem with the new version and advanced workflows. Basically when I use the 9.01 the loading times become very long like 3 hours. I tried various combinations but it seems that it uses the cpu automatically. I have a rtx3060 12 gb

    patchworkpantsDec 31, 2024

    Have you been looking at VRAM usage on Task Manager? The rendering time on generative AI models will get crazy slow if it maxes out and starts using your regular RAM to make up the extra that it needs. Also sometimes your computer might be using some VRAM for other tasks, like some of the tabs you have open in Chrome for example. I always have trouble figuring out which tabs and why. Lol.

    CavernustDec 31, 2024

    @patchworkpants yes, i also tried using comfyui desktop with the internet browser closed. i think the problem is something else. if i use 091 on a basic workflow, everything works. if i use advanced workflows like the official one, but with the same encoder models etc, in that case it goes really slow like 3 hours or even more. i haven't been there to check a lot. i tried all the encoders, fp8, scaled, fp16 but nothing changes.

    the only case in which it worked was with a custom workflow 512x512 and it takes about 4 minutes. i also checked the steps and the other settings and everything is correct. damn

    WilliansJan 5, 2025
    CivitAI

    very bad, I have better results with svd

    theallyJan 5, 2025· 2 reactions

    The img2video is surprisingly good for such a lightweight model, but text2video is a little rough, true.

    941052Jan 14, 2025
    CivitAI

    there is the old v0.9 version here with sha256: 63111264103e4bb8e7f5f7a653e9417d83c84759a48a945f9fabf7f7ea0d9984 while in the huggingface official Lightricks space there is the updated version with sha256: 45369c2ae94e3949b127ef99ca9fe173b077f1ca0fecf8185e4115afcbc09581

    336012Jan 27, 2025
    CivitAI

    Had trouble with movement in image to video, I found using jpg compression to add subtle noise to the image, helps greatly

    theallyJan 27, 2025· 1 reaction

    Yes, in the Education Article linked at the top I speak about this, and provide a workflow to rectify the lack of movement. Glad you got it working!

    336012Jan 27, 2025· 1 reaction

    @theally  great to know about that, I have been making a node for it to work in Invoke AI using the diffuser pipeline so was referencing the pipeline docs

    TheKnightsWhoSayNIFeb 8, 2025
    CivitAI

    Does "Lightricks LTXV" is better than gguf version "ltx-video-2b-v0.9.1-r2-f32.gguf"s ?

    GoronPlusFeb 15, 2025· 7 reactions
    CivitAI

    I'm genuinely confused as to why there aren't more loras for this model.

    VCominosApr 19, 2025

    It's quite hard to make it understand what you want, plus it's really small in size, so it has not so much potential as others like hunyuan or wan2.1

    LG89Mar 6, 2025· 3 reactions
    CivitAI

    @theally FYI, they have released LTXV 0.9.5
    https://huggingface.co/Lightricks/LTX-Video/blob/main/ltx-video-2b-v0.9.5.safetensors

    It's already compatible with ComfyUI (they've also updated the workflows)

    https://comfyanonymous.github.io/ComfyUI_examples/ltxv/