LTX Video

Self-Hostable

Category: video • Released 2024

Realtime open-source video diffusion transformer running faster than playback speed

LTX Video is an open-weights video model created by Lightricks. Designed for high efficiency and speed, it can synthesize high-resolution 24fps video clips in seconds on consumer hardware. Fully compatible with ComfyUI and Hugging Face diffusers.

Pricing Structure & Model AccessOpen-weight availability does not include compute or hosting. Check the current repository license and any provider terms before commercial use.

freeOfficial Pricing

Free Access QuotaLTX Video open weights and code can be run in a compatible local or hosted inference setup; compute is separate.

Free Tier LimitsGeneration speed, model size, and output resolution depend on GPU memory, model configuration, and inference stack.

Requirements & Verification

No Account Required (Instant)Verified: 2026-10-18

Source repository

Core Capabilities & Features

Realtime 24fps video rendering

Open source (Apache 2.0)

ComfyUI integration

Text-to-video

Image-to-video

Who It's Best Suited For

Ideal for practitioners focused on Interactive video prototyping, Real-time creative exploration, Local ComfyUI workflows.

Execution Environment

Self-hostable via Docker or local GPU (macOS, Linux, Windows).Hardware / setup: Nvidia GPU with 8GB–12GB VRAM; ComfyUI or Python diffusers environment