The official LoRA development tool 'LTX Trainer' for the local video generation AI 'LTX-2.3' has been significantly updated, making it easier to create character LoRAs and other content.

A major update to ' LTX Trainer ,' the official LoRA development tool for the video generation AI ' LTX-2.3 ,' was announced on June 17, 2026.
Introducing The New LTX Trainer: One Framework, Every Training Mode | LTX
https://ltx.io/newsroom/introducing-the-new-ltx-trainer-one-framework-every-training-mode
LTX-2/packages/ltx-trainer/README.md at main · Lightricks/LTX-2 · GitHub
https://github.com/Lightricks/LTX-2/blob/main/packages/ltx-trainer/README.md
Most AI video tools stop at prompting, but LTX goes a layer deeper.
— LTX.io (@ltx_io) June 17, 2026
Introducing the new LTX Trainer.
Train LoRAs and IC-LoRAs across video, audio, cross-modal, and reference-conditioned workflows from a single framework.
Plus:
• New agentic skill
• Flexible pricing
•… pic.twitter.com/ee0wVZAcba
LTX-2.3 is a video generation AI that can create videos with voiceovers and background music. It has gained popularity because its AI model is publicly available for free and can be run locally on a personal PC using tools like ComfyUI .
Local video generation AI 'LTX-2.3' released, along with free PC application 'LTX Desktop' - GIGAZINE

LTX-2.3 allows you to control the art style and easily depict specific people or characters using LoRA and IC-LoRA. While LoRA could be created using the official LoRA development tool, LTX Trainer, it previously required building custom code for each application. With this update, you can now create LoRA and IC-LoRA for a variety of purposes, such as 'generating video from text (T2V)', 'generating video from images (I2V)', 'generating video from audio (A2V)', 'generating video from video (V2V)', 'video extension', 'inpainting', and 'outpainting', simply by rewriting a single configuration file.
Additionally, a SKILL.md file has been released to instruct an AI agent to create LoRA data using LTX Trainer, making it easy for even those unfamiliar with video generation AI to create LoRA data by instructing agents such as Claude Code. However, LoRA training using LTX Trainer requires an NVIDIA GPU with 80GB or more of VRAM. Even when using memory-saving features, a GPU with 32GB or more of VRAM, such as a GeForce RTX 5090, is essential.
Several LoRA demo videos created using the LTX Trainer have also been released. Below is a LoRA example that can generate a consistent video by inputting character images.
Ingredients
— LTX.io (@ltx_io) June 17, 2026
Combine characters, props, locations, and styles into fully realized video worlds generated from your reference sheet
(2/10) pic.twitter.com/X3cEFkddTe
LoRA converts videos taken during daylight hours into videos taken at night.
Day → Night
— LTX.io (@ltx_io) June 17, 2026
Transform daytime footage into nighttime scenes
(4/10) pic.twitter.com/rmqViuxg0M
There's also LoRA, which converts black and white videos into color videos.
Colorization
— LTX.io (@ltx_io) June 17, 2026
Bring grayscale footage back to life
(5/10) pic.twitter.com/UMfIxUxb8S
The example LoRA is available at the following link.
LTX-2.3 Creative Lab - a Lightricks Collection
https://huggingface.co/collections/Lightricks/ltx-23-creative-lab
The LTX Trainer documentation can be accessed via the following link. However, at the time of writing, it appears to be under maintenance, and many links are broken.
LTX-2 Training | LTX Documentation
https://docs.ltx.video/open-source-model/ltx-trainer/ltx-training
Related Posts:






