LTX-Video

LTX-Video (Lightricks)

Text-to-video generation focused on speed, control, and practical iteration.

• Video-first by design. Built specifically for text-to-video, not adapted from image diffusion, with an architecture tuned for temporal consistency and motion.

• Fast generation. Optimized for low latency and efficient sampling, making short video generation and rapid iteration practical on a single GPU.

• Controlled motion. Handles camera movement, subject motion, and scene changes more predictably than image-to-video pipelines.

• Developer-oriented. Designed to respond well to structured prompts, enabling easier testing, comparison, and prompt refinement.

• Research-friendly release. Open weights intended for experimentation and evaluation, not a locked commercial API.

Why pick it for Norman AI?

LTX-Video is a strong choice when you want to explore text-to-video without heavy infrastructure. It’s well suited for prototyping motion, testing prompt behavior, and comparing video generation approaches in a controlled, developer-friendly setup.

response = await norman.invoke(
    {
        "model_name": "ltx-video",
        "inputs": [
            {
                "display_title": "Prompt",
                "data": "A cat playing with a ball on mars"
            }
        ]
    }
)