【My process】Building in ComfyUI: Video Upscale + Workflow

LTX 2.3 Step 1. Download the Custom Node from GitHub 1. Open your browser and go to the official project page: https://github.com/Lightricks/ComfyUI... 2. In the upper right area of the page (just above the file list), click the green *Code button*. 3. In the dropdown menu, select the very last option — Download ZIP. 4. Save the archive file to your computer. Step 2. Extract into the ComfyUI Directory 1. Open the downloaded ZIP archive (it will likely be named ComfyUI-LTXVideo-main.zip). 2. Open your file explorer, go to your main ComfyUI folder, and enter the custom_nodes directory. 3. Extract or drag-and-drop the folder from the archive right into custom_nodes. 4. Important Detail: After extracting, the folder will be named ComfyUI-LTXVideo-main. Rename it manually by deleting the -main suffix so it is named exactly: ComfyUI-LTXVideo. Upload model files directly from the Hugging Face platform via your browser. Below are direct instructions and links to the required download pages File 1. Video Model: LTX 2.3 (Distilled FP8) 1. This is the main video generation model, compressed to an 8-bit format to save video memory (VRAM). 2. Go to the Hugging Face repository page: Lightricks/LTX-2.3-fp8 (Fhttps://huggingface.co/Lightricks/LTX...) 3. Locate the file named exactly: ltx-2.3-22b-distilled-fp8.safetensors 4. Where to place it: Move the downloaded file into this exact folder: ComfyUI / models / checkpoints / LTX-Video (create the LTX-Video folder manually if it does not exist) File 2. Text Encoder: Gemma 3 (Heretic v2 FP8) This is the specialized version of Google's text processor shown in your original screenshot. It interprets prompts accurately without false safety triggers. 1. Go to the Hugging Face repository page: https://huggingface.co/DreamFast/gemm... 2. Find the exact filename from your screenshot: gemma-3-12b-it-heretic-v2_fp8_e4m3fn.safetensors 3. Click the Download button next to it. 4. Where to place it: Move the downloaded file into this folder: ComfyUI / models / clip Bonus File: Spatial Upscaler (Fixes Common Errors) Official ComfyUI-LTXVideo workflows often crash on the first run because they require a dedicated latent upscaler model. Downloading this ahead of time prevents that error. 1. Go directly to this Hugging Face link: https://huggingface.co/Lightricks/LTX... 2. Download the file named: ltx-2.3-spatial-upscaler-x2-1.1.safetensors 3. Where to place it: Go to your ComfyUI / models / directory. 4. Create a new folder there named exactly latent_upscale_models, and put this upscaler file inside it. Once all three files are downloaded and installed, launch ComfyUI—the nodes will automatically be added to the drop-down menus! CUSTOM NODES MANAGER: ComfyUI-Manager ComfyUI-LTXVideo ComfyUI-VideoHelperSuite ComfyUI_InstantID[Matteo] PuLID_ComfyUI ComfyUI_IPAdapter_plus comfyui_controlnet_aux ComfyUI Impact Pack ComfyUI-Custom-Scripts ComfyUI_LayerStyle rgthree-comfy ComfyUI-AdvancedLivePortrait ComfyUI-KJNodes ComfyUI-Easy-Use ComfyUI-Crystools ComfyUI-Florence2[Kijai] efficiency-nodes-comfyui ComfyUI-segment-anything-2 comfyui-inpaint-nodes ComfyUI_essentials ComfyUI-Inpaint-CropAndStitch CG Use Everywhere ComfyUI ArtVenture Song: No One Here Gets In Alive - National Sweetheart Never Coming Down - The Soundlings