Descriptions:
Nerdy Rodent demonstrates a comprehensive LTX Video 2 workflow inside ComfyUI, combining text-to-video, image-to-video, first/last-frame interpolation, and audio generation within a single organized node graph. The tutorial covers the full loader setup — including distilled and non-distilled model variants, spatial and temporal latent upscalers, and optional LoRAs for added detail and camera movement — along with practical VRAM guidance for running the 27GB model on a consumer RTX 3090 via system RAM spillover.
Boolean-controlled switches cleanly separate stage-one and stage-two sampling paths, and the presenter explains how audio and video latents are concatenated before being passed to the sampler. The workflow also includes a power LoRA loader for the detail and camera LoRAs, a combo-box model loader (with a noted quirk requiring browser refresh when switching tabs), and a first/last-frame control node from the ComfyUI TTP Tool Set pack that supports an optional middle-frame input for extended interpolations.
LTX Video 2 is source-available under a restrictive commercial license, and the video explicitly recommends reviewing the license file before any deployment use. For ComfyUI users looking to consolidate multiple video generation modes — including audio-synced outputs — into one maintainable workflow, this serves as a thorough starting point.
📺 Source: Nerdy Rodent · Published January 15, 2026
🏷️ Format: Hands On Build
