This ComfyUI workflow applies TheBurgstall’s Googlyeyes LoRA for LTX-Video 2.3 to any input clip, turning ordinary footage into a playful googly‑eyes edit. It ingests your video with VHS_LoadVideo, trims the timeline using skip_first_frames and frame_count, and then routes the frames through the LTX 2.3 inference group (node id: 2e37ecc9-770a-4ac2-972c-e953901f49d3), where the Googlyeyes LoRA is loaded over the LTX‑Video 2.3 base model. Finally, VHS_VideoCombine reassembles the processed frames back into a video at the original frame rate.

Technically, the pipeline is video‑to‑video: frames decoded by VHS_LoadVideo are processed in sequence by the LTX 2.3 node that applies the LoRA to emphasize and transform eye regions while preserving overall scene structure. Because it runs through a video‑aware model pipeline, the effect tends to remain consistent across adjacent frames. The simple set of controls—especially skip_first_frames and frame_count—makes it practical to target only the part of the timeline you want, saving VRAM and time on longer clips.