Watch and track your favorite playlist.
Curated by: VictoryCat (71 videos)
In this video we go over how you can create an AI generated video using LTX 2.3. We will be able to generate videos in either Landscape or Portrait mode, and the videos will have both video and sound. In my opinion this will be better than Sora or other video generation techniques. Coming in at about 1/20th the price of using a video generation API when run on the cloud, and free if run locally. Timestamps: 0:00 - Introduction 1:20 - Video samples 2:09 - Setting up Runpod 3:59 - Storage in Runpod 6:07 - Starting a pod 7:33 - Downloading files on runpod 8:34 - Setting up hugging face token 9:41 - Downloading Checkpoint, runpod 11:44 - Downloading LoRA on Runpod 13:30 - Downloading text encoder 16:14 - Navigating files in runpod 16:40 - All File Locations *important 18:14 - Opening ComfyUI 18:57 - Downloading workflow 19:37 - Downloading Custom nodes for workflow 20:48 - Explanation of workflow 24:48 - Default video generation 26:43 - Modifying prompt 28:22 - Generating a longer video 28:54 - Chef video 30:10 - Generating Video from image 31:24 - Viewing video from image 32:14 - Outro To set up a runpod please use this link: https://runpod.io?ref=9tiy86jm This will give you a $5 bonus for you first ten dollars put on runpod, and also it will help me a little bit, as I seek to make more of these videos for all of you and hopefully start training some models in the near future, which I will release openly. The workflow we will be using is here: https://github.com/Lightricks/ComfyUI-LTXVideo/blob/master/example_workflows/2.3/LTX-2.3_T2V_I2V_Single_Stage_Distilled_Full.json To download the necessary files, use the following links: For main checkpoint: https://huggingface.co/Lightricks/LTX-2.3-fp8/tree/main From here download the ltx-2.3-22b-dev-fp8.safetensors file, and put it in ComfyUI/models/checkpoints/ alternatively you could to to: https://huggingface.co/Lightricks/LTX-2.3/tree/main and download the ltx-2.3-22b-dev.safetensors file, putting it in the same location, but note that this file will be larger and require a large GPU to run. For the LoRA: https://huggingface.co/Lightricks/LTX-2.3/tree/main And download the ltx-2.3-22b-distileled-lora-384-1.1.safetensors and put it in ComfyUI/models/loras/ Finally for the text encoder: https://huggingface.co/Comfy-Org/ltx-2/tree/main/split_files/text_encoders Download the gemma_3_12B_it_fp8_scaled.safetensors and put it in ComfyUI/models/text_encoders/ Note that for this section you can use any of the gemma models on this page, including the fp4 or full verison. The difference will mainly be size and quality. In my experience I have found that the fp8 version works very well, but it is worth experimenting with the other versions. To install the models on runpod, you are going to need to use the following commands: HF_TOKEN="token" This command will create a variable with your hugging face token, make sure to replace token with the actual value of whatever your token is, and to leave the quotation marks. wget --header="Authorization: Bearer=$HF_TOKEN" "download_link" -O filename This command with download the tile from download_link, remember to replace that with the download link, and it will give the file the name of whatever you put after the -O flag. It will use the authorization of the variable you created when you made the HF_TOKEN variable. Make sure you are in the location you want the file before running this command. A pair of good sources from Lightricks for video generation prompts that i have found are here: https://ltx23.github.io/ltx-2-3-best-prompts/ https://ltx.io/model/model-blog/ltx-2-3-prompt-guide If you want more information on ComfyUI, please check out the following videos: https://youtu.be/PeYl2LoZKd0 This is a more general video on how to generate images on ComfyUI in runpod. https://youtu.be/EGy5z3ujzL0 This is how we got ComfyUI running on a small mac mini. Note that this is more for image generation, not for videos. #tutorial #aivideogeneration #comfyui #runpod #cloudai #learnai