LTX-2.5 Pro

LTX-2.5 Pro is the "now make it good" button of Lightricks' hosted video generation - the quality tier that sits above LTX-2.5 Fast in the same cloud lineup. If you've read the Fast node's page, the input list here is near-identical, and that's the deliberate design of tiered model nodes: the widget set stays the same so you can swap the node, keep the prompt and seed, and pay more only when the draft earns a full-quality render. Pro is what you reach for when Fast got the idea right and the execution needs the big model.

The LTX story is worth repeating here because it sets your expectations. Lightricks is the odd company in video: a consumer-app business whose Facetune and LTX Studio revenue funds the LTX open-weights line, and the KB records a genuine open-weights commitment that survived even the 2026 spinoff. The current-generation LTX-2.5 is cloud-hosted rather than downloadable, and the community's sharpest tool for judging it is the model you can run - LTX-2.3, whose quality jump ("better prompt adherence, no more static I2V, portrait resolutions work, better audio") flipped a meaningful slice of Wan users. Pro tiers are the vendor's claim that the newer generation clears that bar with more headroom for detail, motion quality, and complex prompts. Test that claim against your own footage - nobody's benchmarks decide it for you.

The controls match Fast node-for-node: required positivePrompt, referenceImages/frameImages for I2V and frame conditioning, and the genuinely distinctive audio input that drives the video and sets its length from the audio clip. size defaults to 1920x1080, fps is a dropdown, settings.cameraMovement offers cinematic camera presets, and settings.audio generates synchronized audio in the output (default off - enable it for sound). numberResults, the safety group, outputFormat (MP4/WEBM/MOV), and outputQuality round it out, plus advanced_json if you need to pass a duration the widgets don't expose. Output is a native VIDEO.

Install

Standard Runware pack: search Runware in ComfyUI Manager and install, or clone https://github.com/Runware/ComfyUI-Runware into custom_nodes and pip install -r requirements.txt, then restart and add your Runware API key (Settings → Runware API key, RUNWARE_API_KEY, or runware auth login).

The verdict in practice

Use Fast to find the take, Pro to render it - that two-tier loop is the economically sane way to use cloud video, and the per-run cost printed on the node's title bar keeps you honest about which step you're on. The standing caveat for every hosted video model applies: your prompt and reference clips leave your machine through Runware's cloud, so keep proprietary footage on local LTX-2.3, which remains free, private, and - given Lightricks' track record - probably still most of the way there.