Model reference · open weights

LTX-2.5

Video Lightricks Image→video 2 builds Its own licence terms 1.1M dl/mo

LTX-2.5 is an open-weight video model from Lightricks. LTX-2.5-Diffusers (BF16) weighs 165 GB; the smallest configuration that runs it is H100 80 GB.

For self-hosting, compare this build's requirements below with the GPU you plan to rent.

  • LTX-2.5 is an open-world model by Lightricks designed for local execution and fine-tuning, primarily used to generate synchronized video and audio from text, image, and video inputs.
  • The model supports nine languages, including English, German, and Chinese, and features native multishot generation with a custom Gemma 4 12B text encoder.
  • It is distributed under an "other" license and ships as a split pack of components, including a 22B distilled transformer.

Summary of the Lightricks/LTX-2.5 model card, 2026-10-01. The estimate below is for another build of the family.

What it is

Released byLightricks
Released2026-07-23
VRAM165 GB for the weights

What it runs on

Memory and cards for LTX-2.5-Diffusers (BF16)

165 GBweights, file size
76.0 GBbiggest part
537 MBruntime overhead
CardWeightsMemory
RTX 3060 12 GB … L40S 48 GBdoes not fit
A100 80 GBtight (encoders offloaded)78.2 GB
H100 80 GBtight (encoders offloaded)78.1 GB
RTX PRO 6000 Blackwell 96 GBfits (encoders offloaded)93.8 GB
DGX Spark (GB10) 128 GB unifiedfits (encoders offloaded)107 GB
H200 141 GBfits (encoders offloaded)138 GB
B200 180 GBtight176 GB

Builds

Sizes, precisions & builds

BuildParamsPrecisionWeightsSmallest setup
LTX-2.5 ↗ BF16 201 GBB200 180 GB
LTX-2.5-Diffusers (above) ↗ 19.0BBF16 165 GBH100 80 GB

How it works

How video models work

Prompt / imagestart pointTemporal diffusionframes over timeVideoMP4 clipA video model generates a sequence of coherent frames from your prompt or a starting image.

Running it yourself

Run it on a rented GPU

Rent a machine by the hour. ComfyUI is installed on it. Open ComfyUI through the tunnel, then Templates → LTX-2.5: Text to Video: it loads ComfyUI's own build of this model, the files above.

# on your rented machine (the ssh line is on its page in the console)
# gated on Hugging Face: accept the licence at https://huggingface.co/Lightricks/LTX-2.5
# once, then log in here with a Hugging Face token
hf auth login

# get REPO FILE FOLDER: one file into /workspace/models/FOLDER, where ComfyUI loads it from
get() { hf download "$1" "$2" --local-dir /workspace/hf-files && mkdir -p "/workspace/models/$3" && mv "/workspace/hf-files/$2" "/workspace/models/$3/$4"; }

# the files of the template “LTX-2.5: Text to Video” (41.8 GB)
get Lightricks/LTX-2.5 latent_upscale_models/ltx-2.5-latent-spatial-upscaler-x2-bf16-1.0.safetensors latent_upscale_models
get Lightricks/LTX-2.5 diffusion_models/ltx-2.5-22b-distilled-transformer-comfy-int8-convrot.safetensors diffusion_models
get Lightricks/LTX-2.5 vae/ltx-2.5-video-vae-bf16.safetensors vae
get Lightricks/LTX-2.5 vae/ltx-2.5-audio-vae-bf16.safetensors vae
get Lightricks/LTX-2.5 text_encoders/gemma4-12b-with-proj-ltx-2.5-comfy-int8-convrot.safetensors text_encoders
get Comfy-Org/gemma-4 text_encoders/gemma4_e2b_it_int8_convrot.safetensors text_encoders

start-comfyui
Renting a GPU: connect, tunnels, ComfyUI
# on your computer, in a second terminal: ComfyUI in your browser at http://localhost:8188
# HOST, PORT and BOOKING are your machine's, from its page in the console
ssh -o HostKeyAlias=axforge-BOOKING -o StrictHostKeyChecking=accept-new -L 8188:localhost:8188 dev@HOST -p PORT
© 2026 AxForge · EU-hosted AI infrastructure Pricing Docs Trust Privacy Terms