LTX-2.3 - ltx2.3-22b-distilled-lora

NSFW

This is not my model. Don't ask me questions. I don't know the answers.

Go to the official LTX-2.3 HuggingFace repo for answers.

I've only uploaded the model here as a way to tag the videos I generate with it. It will most likely be taken down once the official owners or CivitAI upload their own version, so be warned.

The following is the original README as of time of publishing.

LTX-2.3 is a DiT-based audio-video foundation model designed to generate synchronized video and audio within a single model. It brings together the core building blocks of modern video generation, with open weights and a focus on practical, local execution.

Model Checkpoints

NameNotesltx-2.3-22b-devThe full model, flexible and trainable in bf16ltx-2.3-22b-distilledThe distilled version of the full model, 8 steps, CFG=1ltx-2.3-22b-distilled-lora-384A LoRA version of the distilled model applicable to the full modelltx-2.3-spatial-upscaler-x2-1.0An x2 spatial upscaler for the ltx-2.3 latents, used in multi stage (multiscale) pipelines for higher resolutionltx-2.3-spatial-upscaler-x1.5-1.0An x1.5 spatial upscaler for the ltx-2.3 latents, used in multi stage (multiscale) pipelines for higher resolutionltx-2.3-temporal-upscaler-x2-1.0An x2 temporal upscaler for the ltx-2.3 latents, used in multi stage (multiscale) pipelines for higher FPS

Model Details

Developed by: Lightricks
Model type: Diffusion-based audio-video foundation model
Language(s): English

Online demo

LTX-2.3 is accessible right away via the API Playground.

Run locally

Direct use license

You can use the models - full, distilled, upscalers and any derivatives of the models - for purposes under the license.

Description

8-step lora for both full and FP8 models

FAQ

Comments (13)

lanceshockerMar 7, 2026· 1 reaction

CivitAI

I see the new LTX... however, I have zero clue where to find the audio and video VAE needed to make this work!

clueless_engineer

Author

Mar 7, 2026· 1 reaction

They're baked into the models.

lanceshockerMar 7, 2026· 2 reactions

@clueless_engineer That answers my question then... Have a recommended workflow for this?

MrReclusive666Mar 7, 2026· 1 reaction

@lanceshocker kijai also has released separate ones, i have them in my fp4 variant as they are required for some of the fp4's i've put up, the vae's are in bf16.

randoodMar 7, 2026· 1 reaction

Do LORAs for previous versions work with this?

happylittleteapotMar 7, 2026· 1 reaction

@lanceshocker the ltx comfy nodes github has example workflows

MrReclusive666Mar 7, 2026· 1 reaction

@randood somewhat varied success atm, some do, some don't, i've been playing with block editing to see if I can make it more consistent.

DarkEngine2024Mar 9, 2026· 1 reaction

@lanceshocker https://www.reddit.com/r/StableDiffusion/comments/1rn3fjv/for_ltx2_use_triple_stage_sampling/?sort=new Three-stage method workflows have apparently given good results, though similar to Wan 2.2's Three-chain setup, there's current debates as to what it actually affects and how much.

MrReclusive666Mar 9, 2026· 4 reactions

@DarkEngine2024 DUUUUUDE, thank you, been playing with this, did some adjustments, with the sigma's, and locked audio latent after base, and my outputs are getting insane, just tweaking some final things.
for anyone else who plays with this, my current settings that ill be posting examples of soon.
base euler ancestral + sgm uniform @ 10 steps
2x sigma with euler ancestral: 0.91218, 0.909375, 0.725, 0.421875, 0.0
4x sigma with euler ancestral: 0.909375, 0.725, 0.421875, 0.0

DarkEngine2024Mar 10, 2026· 1 reaction

@MrReclusive666 Oh really? Thanks a lot, will try these.

randoodMar 7, 2026· 1 reaction

CivitAI

Do existing LTX2 LORAs work?

happylittleteapotMar 7, 2026· 1 reaction

yes, but maybe at different strengths. From my tests the I2V has been stiffer, but raising weights can help. It'll be like Wan2.2, although it doesn't look as intense.

aomyouza2543Mar 9, 2026· 2 reactions

CivitAI

@AiMetatron @DigitalPastel @taek75799 @darksidewalker @FX_FeiHou

Checkpoint

LTXV 2.3

by clueless_engineer

Download (Beta) View on CivitAI

Details

Downloads

1,316

Platform

CivitAI

Platform Status

Available

Created

3/10/2026

Updated

7/26/2026

Deleted

Files

ltx23_ltx2322bDistilled.safetensors

Size:

7.08 GB

SHA256:

2943ab994f3c9d88052e5a2a34cca14e4a2dfc36b1d8c407931d52d5c25dd72b