BYOKey LTXV Text to Video

LTX is Lightricks' video line, and it has two very different halves. The open half - LTX-Video - is famous in this community as the fast local video model, the one that runs in near real time on a single GPU, and the KB's ecosystem guide covers it in depth. The frontier half - LTX-2 - is Lightricks' hosted API model, the professional-quality one you can't download. BYOKey_LTXV_Text2Video is the node that calls that hosted API (api.ltx.video/v1) with your own key and returns a ComfyUI VIDEO. Same family, different world: no local GPU work, just your key against Lightricks' servers.

What it buys you is the thing local LTX-Video can't: resolution up to 3840x2160, longer durations, and AI-generated audio matched to the scene. For a polished, near-cinematic text-to-video pass without renting serious hardware, that's a real capability jump.

How it works

Unlike the pack's poll-based video nodes, this one is a single synchronous call - the tooltip on max_wait says it explicitly: "Maximum seconds to wait for the synchronous response." You POST your request and the node holds the workflow open until Lightricks finishes and hands back the video, which gets wrapped into ComfyUI's VIDEO output type. Simpler than submit-and-poll, at the cost of the node sitting there during generation.

The inputs that matter

• model - LTX-2 (Pro) (default) or LTX-2 (Fast). Pro for finals, Fast for drafts and iteration. The quality gap is exactly what you'd expect from the names.
• duration - 6 to 20 seconds (default 8), in even steps. This is the longest duration in the pack's video nodes.
• resolution - 1920x1080 (default), 2560x1440, or 3840x2160. 4K text-to-video is still a "wow, it exists" feature, but each step up costs more and takes longer - 1080p is the sane default.
• fps - 25 or 50. 50fps is for the footage that wants smooth motion; 25 is standard and cheaper.
• generate_audio - the feature to care about: when true, the video includes AI-generated audio matched to the scene. This is the capability the KB flags as genuinely rare - Veo 3 and a handful of others do native video audio, and LTX-2 is in that club. Turn it on for anything you're not going to re-score yourself.

Output is video (VIDEO) - feed it to a video preview or save node to see it.

Installing it

It's one of the 37 nodes in comfyui-byokey, with zero extra dependencies beyond ComfyUI's bundled torch/aiohttp/PIL/numpy/av. Install via ComfyUI Manager (search "byokey") or:

cd ComfyUI/custom_nodes
git clone https://github.com/MeteorAndy/comfyui-byokey.git

Restart, and it's under api/byokey/video. No model downloads - Lightricks does the compute.

Gotchas

Two things to keep in mind. First, the synchronous wait: there's no polling here, so a busy Lightricks queue means the node just blocks. If it times out, raise max_wait (up to 3600s) rather than assuming it's broken. Second, cost scales hard with the top-end settings - 20s at 4K with audio is a very different bill than 8s at 1080p without it, so treat the big settings as final-pass-only. And a good prompt matters more here than anywhere in the pack: LTX-Video's whole reputation is that it's prompt-faithful but punishes sloppy wording, and the hosted LTX-2 carries the same DNA - spend the extra minute writing a proper shot-and-motion description. Standard BYOKey hygiene: your key is saved into the workflow JSON, so scrub before sharing workflows. For 4K, audio-capable video from a prompt, this is the node.