OpenArt LTX‑2.3 Launch

OpenArt has added LTX‑2.3 to its video tools, bringing a newer open model for text-to-video and image-to-video generation. (x.com) The underlying model comes from Lightricks, which updated its LTX‑2.3 collection on Hugging Face in the past day and lists both a standard release and an fp8 variant. The LTX‑2 GitHub repository now tells users to download LTX‑2.3 checkpoints, including 22 billion-parameter dev and distilled files. (huggingface.co) (github.com) Video models turn text, images, and sometimes audio into short clips by predicting one frame after another, like autocomplete for motion. LTX’s documentation says its system generates audio and video together in one pass, instead of stitching sound on afterward. (docs.ltx.video) Lightricks made LTX‑2.3 the default model in its application programming interface on March 5, 2026, and said the update brought sharper fine details and cleaner audio. The same changelog says the release came in two versions: Pro for the full feature set and Fast for quicker text-to-video and image-to-video jobs. (docs.ltx.video) The current LTX model page also says LTX‑2.3 supports portrait and landscape video up to 4K, cinematic 24 or 48 frame-per-second output, and first-to-last-frame control for image-to-video. That portrait support is part of why OpenArt is highlighting the model for creator workflows built around vertical video. (docs.ltx.video) OpenArt’s own LTX video page already offers 1080p, 2K, and 4K settings, plus optional audio generation, showing how the company is packaging outside model releases inside its broader creator suite. OpenArt’s homepage says that suite mixes models from several providers, including Google Veo, Kling, Nano Banana Pro, and Seedance. (openart.ai 1) (openart.ai 2) LTX is also being positioned as an open alternative to closed video systems. Lightricks’ GitHub page describes LTX‑2 as an open-access audio-video model with application programming interface access, local inference support, ComfyUI integration, and LoRA training tools for customization. (github.com) That gives OpenArt a way to offer a model that can appeal both to casual users on its web app and to developers who want the same family of models in local or custom pipelines. LTX’s quick-start guide recommends ComfyUI for local generation and also offers a desktop app with a built-in editor. (docs.ltx.video) The release lands as AI video platforms compete on the same checklist: sharper frames, better prompt following, more stable motion, and cleaner audio. OpenArt’s pitch is that LTX‑2.3 now covers those basics inside a tool built for fast publishing, especially in vertical formats. (x.com) (docs.ltx.video)