LTX-2
LTX-2 by Lightricks: an open model that makes video together with its sound, for text to video and image to video.
What it is
LTX-2 generates the picture and a matching soundtrack (speech, effects, music) in one run. It follows the earlier LTX-Video line, which was known for speed, but its files are a new family. Its templates include a two-stage approach with its own upscale models for higher resolution.
Files and nodes
LTX-2 uses its own model, text encoder and latent nodes, plus upscale models; start from ComfyUI's official LTX-2 templates for the exact files and nodes rather than building by hand.
Settings that work
Take size, length (8n + 1 frames), frame rate (around 24 to 25), steps and guidance from the template. Prompt: detailed sentences that describe the scene, the action, the camera and what is heard.
What fits
Only LTX-2 LoRAs. LoRAs for the older LTX-Video (0.9.x) models are not compatible.
How the app helps
Shown as LTX-2 on model cards. The lesson Video with sound: LTX-2.5 builds a clip with sound in two stages.