LTX-2

an audio-visual base model based on the DiT architecture, developed for synchronized generation of video and audio within a single model. It incorporates key components of modern video generation systems, including open weights and optimization for local use.

Key Features:

• Audiovisual content generation based on text, images, video, or audio.
• Supported tasks: Image-to-Video, Text-to-Video, Video-to-Video, Audio-to-Video, Text-to-Audio, and others.
• Built-in audio-video synchronization.
• Limitations:
  • Not designed to provide factual information.
  • Frame size must be divisible by 32; number of frames must be divisible by 8+1.
  • Audio without speech may have low quality.
  • Prompt execution accuracy depends on the writing style. For details, see the Prompting guide.
  • The model is distributed with a warning about the potential generation of inappropriate or offensive content.

---

The model is a component of the video generation pipeline, consisting of:

• audio VAE: ~153M parameters, 
• connectors: ~1.4B parameters, 
• latent upsampler: ~497M parameters, 
• Text encoder: ~12B parameters,
• Transformer: ~19B parameters,
• VAE: ~1.2B parameters, 
• vocoder: ~56M parameters

Total: ~34B parameters

---

For local running, NVIDIA recommends using 24GB+ GPU to generate a 4-second video at 720p24 resolution (with 20 steps).

Announce Date: 06.01.2026
Parameters: 19B
Developer: Lightricks
Diffusers Version: 0.37.0.dev0
License: LTX-2 Community License Agreement

Use our pre-built public endpoints for free to test inference and explore LTX-2 capabilities. You can obtain an API access token on the token management page after registration and verification.

Rent your own physically dedicated instance with hourly or long-term monthly billing.

We recommend deploying private instances in the following scenarios:

• maximize endpoint performance,
• enable full context for long sequences,
• ensure top-tier security for data processing in an isolated, dedicated environment,
• use custom weights, such as fine-tuned models or LoRA adapters.

Contact our dedicated neural networks support team at nn@immers.cloud or send your request to the sales department at sale@immers.cloud.