LTXV Distilled 8 steps

LTXV Distilled 8 steps is a ComfyUI workflow for image-to-video generation: it accepts an image-to-video prompt and optionally an input image, then produces video using base or tiled sampling and latent upscaling.

It offers an LTXV base sampler and a tiled sampler branch for higher resolutions and extended generation scenarios.

Optional image conditioning supplies visual input anchors, and AdaIN blending can use reference latents during latent upscaling.

Standard or tiled VAE decoding feeds VHS Video Combine for direct video output.

Provide an image-to-video prompt and, optionally, an input image for visual conditioning.

The workflow can produce video through a low-res base path or a tiled-and-upscaled path, with optional Film Grain & Resize finishing.

It uses the LTXV base model and is published by Ltxvideo.

The workflow requires the ComfyUI-KJNodes node pack.

The listed model files are ltxv-13b-0.9.7-distilled.safetensors, ltxv-spatial-upscaler-0.9.7.safetensors, and t5xxl_fp16.safetensors.

For higher-resolution or extended-generation scenarios, use the tiled sampler with latent upscaling; for the low-res path, use the base sampler.

Suggestion · non vérifié

The base sampler supports sigma, cfg, and stg tuning, and Film Grain & Resize is optional.

The workflow is described as compatible with LoRA and conditioning workflows.

Before use, verify that these model files are available under the exact filenames listed above.

Suggestion · non vérifié

Before use, verify that ComfyUI-KJNodes is installed.

Suggestion · non vérifié