Film Auteur (LTXV) 2.0.5 turns ComfyUI into a one-node video production suite
Infinite-length clips, audio-to-video, and director mode in a single ComfyUI node.
triXope's Film Auteur (LTXV) custom node for ComfyUI has been updated to version 2.0.5, bringing a comprehensive set of video generation and editing features into a single interface. Designed for LTX 2.3, the node eliminates the need for separate video extenders or multiple runs by handling any number of prompts—each separated by "|" —and generating clips of infinite length through autoregressive chunking with sliding context windows. Users can input text, images (including batches for storyboarding), audio, or even reference images with IC-LoRA guide for consistency.
The update includes specialized features like Director Mode for multi-shot inferencing, normalized attention guidance (NAG), Ollama integration for prompt enhancement, and upscaling options (1–2 spatial passes plus temporal upscaling that doubles framerate and improves motion/lip sync). It also offers face restoration, an integrated audio mastering pass (soft limiter and normalization), built-in sageattention and fp16 accumulation, chunk feed-forward for efficiency, and unload model/cache switches. Planned additions include prompt relay keyframes and RTX super resolution. Despite being a hobbyist project, the node demonstrates production-level polish and is available via ComfyUI Manager or GitHub.
- Single node handles text-to-video, image-to-video, audio-to-video, and reference-image guidance with IC-LoRA.
- Infinite-length generation via autoregressive chunking and sliding context windows, preventing OOM errors.
- Includes Director Mode for multi-shot sequences, spatial/temporal upscaling, face restoration, and Ollama prompt enhancement.
Why It Matters
Simplifies LTX video workflows into one node, enabling professionals to generate long, multi-shot clips without complex pipelines.