Support Center

LTX-2 is a DiT-based audio-video foundation model that generates synchronized video and audio in a single unified process. Unlike other models that generate video and audio separately, LTX-2 creates them together with natural timing and synchronization. It supports 4K resolution at 50 FPS and includes LoRA customization capabilities.

For optimal performance, we recommend:

• Recommended: RTX 40 Series or newer with 16GB+ VRAM
• Optimal: 24GB+ VRAM (RTX 3090/4090) for 720p 24fps 4-second clips
• Minimum: 8GB VRAM for 540p at reduced duration

Alternatively, use the online playground which requires no local GPU.

LTX-2 can generate videos up to 20 seconds in a single generation. The LTX Platform currently supports 6, 8, or 10-second clips, with 15-second support coming soon. For longer content, you can chain multiple generations together in your editing workflow.

Yes, subject to our Terms of Service, you may use AI-generated content for personal and commercial purposes. However, please review the prohibited uses section and ensure your use case complies with all applicable laws and regulations.

The LTX-2 base (dev) model is fully trainable. You can create custom LoRAs for style, motion, or identity in under an hour using the LTX-2 Trainer package. Visit our GitHub repository for detailed training instructions and example workflows.

Yes! LTX-2 has built-in support for ComfyUI with native nodes available in ComfyUI Manager. NVIDIA provides a detailed quick-start guide for running LTX-2 in ComfyUI, including optimized workflows for different GPU configurations.

Audio quality can vary based on the prompt. LTX-2 tends to produce better audio quality when the prompt includes speech or dialogue. Non-speech audio (ambient sounds, music) may have lower quality. For best results, include specific audio descriptions in your prompts.

Please use our GitHub Issues page to report bugs or request new features. Before submitting, please check if a similar issue already exists.