LTXV Text To Video

The LTXV Text To Video node (class LtxvApiTextToVideo) is Lightricks' text-to-video API node that ships in ComfyUI core. It runs the LTX-2 generation models on Lightricks' servers through your Comfy account - LTX-2 (Pro) for quality, LTX-2 (Fast) for the distilled speed model - and hands you back a finished video with no local GPU work and no model downloads. It's billed per generation to Comfy account credits, like every partner node.

The mechanism is straightforward: you pick a model, write a prompt, choose duration/resolution/fps, and the node calls Lightricks' endpoint, then returns the video. The unusual part for an API node is that it's synchronous - it downloads the finished file directly rather than returning a generation id, so you just wait and get a VIDEO out. And it can go big: up to 20 seconds long, and resolutions up to 3840×2160 (4K), at 25 or 50 fps.

The inputs that matter:

• model - LTX-2 (Pro) or LTX-2 (Fast). Fast is the distilled model: quicker and cheaper-feeling, Pro is the quality pick.
• prompt - describe the whole scene. Lightricks' own guidance for LTX-2 is to write one flowing paragraph covering shot, action, characters, and camera - LTX-2 quality tanks if you prompt it like Stable Diffusion.
• duration - 6 to 20 seconds. Caveat from the source: anything over 10s is only allowed for the Fast model at 1920×1080 and 25 fps.
• resolution - 1920×1080, 2560×1440, or 3840×2160.
• fps - 25 or 50.
• generate_audio - optional, default off. Turn it on to get AI-generated audio matching the scene, which is LTX-2's signature trick (it's one of the first open-weights-family models to do synchronized video+audio).

Single VIDEO output into Save Video.

Here's the thing you need to know before you build anything on it: this node is deprecated. ComfyUI marks it (and the sibling LtxvApiImageToVideo) as such, and the current, supported LTX API path is the LTX-2.5 nodes that landed in mid-2026 - Ltx25TextToVideoNode and friends, which add proper multi-shot scenes and the Fast/Pro tiering. The node still works, but new work should go to the 2.5 line.

Two practical notes. First, "LTXV" here is the API family, not the local open-weights LTX-Video models you download and run on your own GPU - those are a separate set of core ComfyUI nodes with totally different (free, local) mechanics. Don't confuse the two when Googling. Second, partner-node house rules apply: log into your Comfy account or it won't run, and --disable-api-nodes hides the whole API family from your node search if you want it gone.