LTX Video Model in ComfyUI - Complete Install Guide

Getting Started with ComfyUI Manager

First, open ComfyUI and navigate to the manager section. Scroll down to verify you have a recent version, typically no more than a few days old. If your version needs updating, click the update all button. Should this method fail, locate the update folder and execute the update ComfyUI BAT file.

Setting Up Your Workflow

Begin by uploading a modified workflow based on recommended configurations. Three workflow versions are available for different aspect ratios: landscape, square, and portrait. You can test each to determine which performs best for your needs. All necessary resources are provided in a comprehensive Pixaroma note for easy access.

Installing the Required Models

Download the FP8 distilled model version and place it in the checkpoints folder. Navigate to your ComfyUI folder, then to the models folder, and locate the checkpoints folder for proper placement. The model will load automatically from this location.

For the clip model, choose the appropriate size based on your available memory. Place this model in the text encoders folder and load it through the designated node. An upscaler model should be placed in the upscale models folder for enhanced output quality.

Installing Essential Custom Nodes

Access the manager and navigate to the custom nodes manager. Install the following required nodes:

- Video Helper - Search and install this node for video processing capabilities
- Easy Use - Install this node for simplified operations
- RG3 - Add this node for additional functionality
- LTX - Install the specific LTX node for model compatibility
- KJ Nodes - The final required node for complete functionality

After installing all nodes, click the restart button and wait for the application to restart completely.

Understanding the Workflow Structure

The workflow consists of two main components. The first part generates video from a static image using the LTX Workflow FP8 version for image-to-video conversion. Input an image at the specified size to receive a video of matching dimensions. The second part utilizes an upscaler to double the output size for enhanced quality.

Configuring Image Input and Prompts

Upload your image, keeping in mind this setup works with 3:2 ratio images. Different ratios will be cropped automatically, though you can modify settings for 16:9 ratios. Create a detailed prompt describing the subject, movement, camera work, and desired effects. Green-colored nodes indicate editable parameters, while others can remain at default settings.

Optimizing Base Sampler Settings

Configure width and height using multiples of 32 or 64 for optimal results. Best performance occurs with sizes under 1280 by 720 pixels. Uploaded images resize automatically to these dimensions, with crop options available to prevent distortion.

Adjust the CRF value to control motion perception and fluidity. Higher CRF values add subtle noise that reduces static frames and creates more natural motion. Values between 20 and 30 are recommended for optimal motion perception in output videos.

Setting Frame Rate and Duration

Frame count determines video length based on your chosen frame rate. Calculate total frames by multiplying frames per second by desired duration, then add one extra frame. For example, 24 fps multiplied by 4 seconds equals 96, plus 1 totals 97 frames.

Set frame rate in both the conditioning node and video combine node for consistency. Longer videos may contain more artifacts and require significantly more processing time due to increased frame counts.

Be a Filmmaker Today

This tutorial provides everything needed to start creating professional-quality videos using the LTX model in ComfyUI. The new distilled version offers improved speed and quality, making local video generation more accessible than ever. Experiment with different settings and share your discoveries with the community.

Level up your team's AI usage—collaborate with Promptus. Be a creator at https://www.promptus.ai