LTX-2.3 Manual Install Guide

Manual Install Method from my last Youtube video: https://youtu.be/nOCMsqVujBI



Windows Installation (Local PC)
Step 1: Prerequisites

Make sure you have these installed before doing anything else:

7-Zip
https://www.7-zip.org/

Git for Windows
https://git-scm.com/download/win

Reboot once after installing Git to ensure it is available in cmd.

Step 2: Download and Extract ComfyUI

Download the official NVIDIA portable build of ComfyUI:

ComfyUI v0.18.0 (NVIDIA Portable)
https://github.com/comfyanonymous/ComfyUI/releases/download/v0.18.0/ComfyUI_windows_portable_nvidia.7z

Extract it with 7-Zip.

Your folder structure must look like this:

ComfyUI_windows_portable/
├── ComfyUI/
│ ├── models/
│ └── custom_nodes/
├── python_embeded/
└── run_nvidia_gpu.bat

If this structure is wrong, stop here and fix it.

Step 3: Install Required Custom Nodes

Open Command Prompt inside:

ComfyUI_windows_portable\ComfyUI\custom_nodes

Then clone all required nodes:

git clone https://github.com/ltdrdata/ComfyUI-Manager.git
git clone https://github.com/city96/ComfyUI-GGUF
git clone https://github.com/rgthree/rgthree-comfy
git clone https://github.com/yolain/ComfyUI-Easy-Use
git clone https://github.com/kijai/ComfyUI-KJNodes
git clone https://github.com/ClownsharkBatwing/RES4LYF
git clone https://github.com/Lightricks/ComfyUI-LTXVideo
git clone https://github.com/pythongosssss/ComfyUI-Custom-Scripts
git clone https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite
git clone https://github.com/kijai/ComfyUI-WanVideoWrapper
git clone https://github.com/ltdrdata/ComfyUI-Impact-Pack
git clone https://github.com/TTPlanetPig/Comfyui_TTP_Toolset
git clone https://github.com/evanspearman/ComfyMath

Step 4: Install Node Dependencies

Still inside custom_nodes, install requirements using ComfyUI’s embedded Python:

..\..\python_embeded\python.exe -m pip install -U pip

Then, for every node that has a requirements.txt, run:

..\..\python_embeded\python.exe -m pip install -r ComfyUI-Manager\requirements.txt
..\..\python_embeded\python.exe -m pip install -r ComfyUI-GGUF\requirements.txt
..\..\python_embeded\python.exe -m pip install -r rgthree-comfy\requirements.txt
..\..\python_embeded\python.exe -m pip install -r ComfyUI-Easy-Use\requirements.txt
..\..\python_embeded\python.exe -m pip install -r ComfyUI-KJNodes\requirements.txt
..\..\python_embeded\python.exe -m pip install -r RES4LYF\requirements.txt
..\..\python_embeded\python.exe -m pip install -r ComfyUI-LTXVideo\requirements.txt
..\..\python_embeded\python.exe -m pip install -r ComfyUI-Custom-Scripts\requirements.txt
..\..\python_embeded\python.exe -m pip install -r ComfyUI-VideoHelperSuite\requirements.txt
..\..\python_embeded\python.exe -m pip install -r ComfyUI-WanVideoWrapper\requirements.txt
..\..\python_embeded\python.exe -m pip install -r ComfyUI-Impact-Pack\requirements.txt
..\..\python_embeded\python.exe -m pip install -r Comfyui_TTP_Toolset\requirements.txt
..\..\python_embeded\python.exe -m pip install -r ComfyMath\requirements.txt

Yes, this takes time. That’s why the installer exists 😄

Step 5: Download LTX-2.3 Models

All model files are hosted here:
👉 Hugging Face
https://huggingface.co/Aitrepreneur/FLX/tree/main

Place them exactly like this:

models/text_encoders/
ltx-2.3_text_projection_bf16.safetensors
gemma_3_12B_it_fp4_mixed.safetensors

models/vae/
LTX23_audio_vae_bf16.safetensors
LTX23_video_vae_bf16.safetensors

models/unet/

Choose one based on your GPU:

Less than 12 GB VRAM → ltx-2.3-22b-dev-Q5_K_S.gguf

12 to 16 GB VRAM → ltx-2.3-22b-dev-Q6_K.gguf

24 GB+ → ltx-2.3-22b-dev-Q8_0.gguf

models/latent_upscale_models/
ltx-2.3-spatial-upscaler-x2-1.0.safetensors

models/loras/
ltx-2.3-22b-distilled-lora-384.safetensors
ltx-2-19b-ic-lora-detailer.safetensors

Step 6: Optional LTX-2.3 Launchers

If you want the same launcher files as my installer, place these in:

ComfyUI_windows_portable/

Files:
run_nvidia_gpu-LTX2-8GB.bat
run_nvidia_gpu-LTX2-12GB.bat
run_nvidia_gpu-LTX2-16GB.bat

You can get them here:
https://huggingface.co/Aitrepreneur/FLX/tree/main

Step 7: Run ComfyUI

From ComfyUI_windows_portable:

Double-click run_nvidia_gpu.bat

Open browser at http://127.0.0.1:8188

If it launches, you’re done.








RunPod Installation (Cloud)
Step 1: Create Account

👉 RunPod
https://bit.ly/runpodAi

Step 2: Create Pod

Choose a 24 GB VRAM GPU or higher.

Use a ComfyUI template.

Expose port 8188.

Step 3: Inside the Pod

SSH or Web Terminal into the pod.

Make sure you are inside the ComfyUI root where models/ and custom_nodes/ exist.

Step 4: Create Virtual Environment

python3 -m venv venv
source venv/bin/activate
pip install -U pip setuptools wheel

Step 5: Install Torch (CUDA)

pip install torch==2.4.0+cu121 torchvision==0.19.0+cu121 torchaudio==2.4.0+cu121 \
  --index-url https://download.pytorch.org/whl/cu121

Step 6: Install Core Dependencies

pip uninstall -y opencv-python || true
pip install opencv-python-headless==4.12.0.88
pip install "pillow>=11.0.0"

pip install \
  boto3 \
  rotary-embedding-torch \
  deepdiff \
  py-cpuinfo \
  diffusers \
  gguf \
  piexif

Step 7: Clone Nodes

Inside custom_nodes/, clone the exact same nodes as in the Windows section:

git clone https://github.com/ltdrdata/ComfyUI-Manager.git
git clone https://github.com/city96/ComfyUI-GGUF
git clone https://github.com/rgthree/rgthree-comfy
git clone https://github.com/yolain/ComfyUI-Easy-Use
git clone https://github.com/kijai/ComfyUI-KJNodes
git clone https://github.com/ClownsharkBatwing/RES4LYF
git clone https://github.com/Lightricks/ComfyUI-LTXVideo
git clone https://github.com/pythongosssss/ComfyUI-Custom-Scripts
git clone https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite
git clone https://github.com/kijai/ComfyUI-WanVideoWrapper
git clone https://github.com/ltdrdata/ComfyUI-Impact-Pack
git clone https://github.com/TTPlanetPig/Comfyui_TTP_Toolset
git clone https://github.com/evanspearman/ComfyMath

Step 8: Patch Fragile Requirements

These two small patches help avoid common dependency conflicts on RunPod:

sed -i 's@^git+https://github.com/facebookresearch/sam2.*@# sam2 disabled for CUDA/Torch stability@' custom_nodes/ComfyUI-Impact-Pack/requirements.txt || true
sed -i 's/^matrix-client==0\.4\.0/# matrix-client disabled by manual install/' custom_nodes/ComfyUI-Manager/requirements.txt || true

Step 9: Install Node Dependencies

Run:

pip install -r custom_nodes/ComfyUI-Manager/requirements.txt
pip install -r custom_nodes/ComfyUI-GGUF/requirements.txt
pip install -r custom_nodes/rgthree-comfy/requirements.txt
pip install -r custom_nodes/ComfyUI-Easy-Use/requirements.txt
pip install -r custom_nodes/ComfyUI-KJNodes/requirements.txt
pip install -r custom_nodes/RES4LYF/requirements.txt
pip install -r custom_nodes/ComfyUI-LTXVideo/requirements.txt
pip install -r custom_nodes/ComfyUI-Custom-Scripts/requirements.txt
pip install -r custom_nodes/ComfyUI-VideoHelperSuite/requirements.txt
pip install -r custom_nodes/ComfyUI-WanVideoWrapper/requirements.txt
pip install -r custom_nodes/ComfyUI-Impact-Pack/requirements.txt
pip install -r custom_nodes/Comfyui_TTP_Toolset/requirements.txt
pip install -r custom_nodes/ComfyMath/requirements.txt

If RES4LYF gives issues, reinstall it once more manually from inside its own folder:

cd custom_nodes/RES4LYF
pip install --upgrade --force-reinstall -r requirements.txt
cd ../..

Step 10: Download Models

Use curl:

curl -L -o filename "HUGGINGFACE_URL"

Follow the exact same folder structure as the Windows guide.

These are the files you need:

models/text_encoders/
ltx-2.3_text_projection_bf16.safetensors
gemma_3_12B_it_fp4_mixed.safetensors

models/vae/
LTX23_audio_vae_bf16.safetensors
LTX23_video_vae_bf16.safetensors

models/unet/
ltx-2.3-22b-dev-Q5_K_S.gguf
or
ltx-2.3-22b-dev-Q6_K.gguf
or
ltx-2.3-22b-dev-Q8_0.gguf

models/latent_upscale_models/
ltx-2.3-spatial-upscaler-x2-1.0.safetensors

models/loras/
ltx-2.3-22b-distilled-lora-384.safetensors
ltx-2-19b-ic-lora-detailer.safetensors

Step 11: Run ComfyUI

python main.py

Open browser at http://127.0.0.1:8188

If it launches, you’re done.

IF YOU GET OOM error:
EXTRA_ARGS --reserve-vram 10 --cache-none