LTX has released tools that empower creators to train custom AI video models in their own distinctive style using LTX Trainer.

The Trainer lets users fine-tune models on their own data locally, safeguarding intellectual property by avoiding any uploads to external servers.

LTX-2, the company’s second-generation model with Nvidia, is designed for on-device run-time and comes with open-weight access to boost transparency.

The tool aims to lower cost and time barriers, enabling individuals to train capable models without lengthy development or huge budgets, and it includes an agentic assistant to simplify natural language-based creation.

Updates expand training to audio-only and joint audio-video modalities, enabling prompts across modalities like audio-to-video or image-to-video.

Users can customize outputs with adapters called LoRAs and IC-LoRAs to consistently replicate a creator’s style across generated content.

LTX emphasizes protecting creators’ IP and keeping data on the creator’s side, addressing concerns about unauthorized use of artists’ styles in AI.