cycomodel

Updated April 3, 2026:

Official version 1.0.3 has been released, significantly reducing video memory usage. Now, graphics cards with 12GB or more of video memory can run the program. Our tests show that, in a 10-second 720p frame test, the maximum video memory usage is only 13GB! Download Link: https://github.com/Lightricks/LTX-Desktop/releases/tag/v1.0.3

---

April 2, 2026 Update:

1. Added LoRA functionality (place LoRA in the loras folder within the model directory).（Quick test LoRa: https://civitai.com/models/2482513/ltx23）

2. Added model selection capability (currently testing quantization to reduce GPU memory usage; modifying the model does not currently lower the GPU memory requirement, pending future updates).

3. Added multi-frame insertion functionality, with two generation modes: Mode 1: Inserts multiple frames into a latent space to directly generate a long video. Mode 2: Generates many independent first and last frame segments, which are then stitched together to form a complete video.

---

This program mainly optimizes the desktop version of LTX, breaking the generation time limitations and lowering the barrier to use. It now only requires 24GB to run, whereas the desktop version needs 32GB.

Compared to the messy and complex workflows and error-prone nodes in the ComfyUI version, this one integrates all features, including image-to-video, text-to-video, start/end frames, lip-sync, video enhancement, and image generation.

No need to install any third-party software—just install the LTX desktop version and you’re good to go. It’s very simple and efficient.

Tutorial: https://youtu.be/rM_wUogtrOU

Desktop version software download address: https://ltx.io/ltx-desktop

The GitHub version has been uploaded: https://github.com/hero8152/LTX2.3-Multifunctional/tree/main

It can be accessed via ComfyUI, node address: https://github.com/supart/ComfyUI_TY_LTX_Desktop_Bridge