LTX 2.5 / createVideo

Here's the thing that makes the LTX 2.5 nodes feel different from the flux and grok ones in this pack: they're not thin "prompt + a few presets" wrappers. LTX 2.5 is an open-weights model family, and Civitai runs it through real Comfy workflows on its own fleet - so these nodes expose the full Comfy parameter surface: width, height, fps, steps, guidance, seed, LoRAs, even a diffusion-model override. You get local-style control, but the GPU is someone else's and the bill lands in Buzz. createVideo is the general-purpose entry: text-to-video, or image-to-video if you wire a still into images.

It's part of civitai-comfy-nodes, Civitai's official ~160-node pack generated from its Orchestration API spec. Cloud job: submit a videoGen, long-poll, get a native VIDEO tensor back.

The LTX parameter surface

The ones a beginner actually sets:

• prompt (required) and negative_prompt - LTX rewards explicit camera and quality keywords; the community's been saying that since the original LTX-2 launch.
• duration - 3 to 20 seconds (tooltip), default 5. Combined with fps (24 default) and width/height (1280×720 default) it defines the job.
• seed - set it to reproduce a result, leave 0 for random.
• guidance_scale (4) and steps (20) - the standard diffusion dials.
• generate_audio - on by default. LTX's synchronized audio is the family's signature; it's the one open model that does native audio.
• model - enum, default 22b-dev.
• loras - a CIVITAI_LORAS socket; wire a Civitai LoRA Selector to stack community LoRAs.
• diffusion_model - CIVITAI_AIR socket to point the job at a community fine-tune checkpoint (the tooltip name-drops SulphurAI/Sulphur-2-base) while keeping the stock text encoder, VAEs and upscale-LoRA behavior.
• quantity - generate several videos in one job, each with a distinct seed (Seed + slotIndex).
• images - optional IMAGE socket; wire a still to turn this into image-to-video.

Outputs: video (VIDEO), draft_cache_audio (AUDIO), draft_cache_video (VIDEO), workflow_id and raw_json (STRING).

Install and auth

Standard pack install - official, Comfy Registry:

# ComfyUI Manager: search "Civitai Comfy Nodes"
cd ComfyUI/custom_nodes
git clone https://github.com/civitai/civitai-comfy-nodes.git
pip install -r civitai-comfy-nodes/requirements.txt

Restart, authenticate via a Civitai Auth node, CIVITAI_API_TOKEN, or sidebar OAuth.

Common issues

Bigger isn't smarter. The 1280×720 default is there for a reason; pushing width/height up on a cloud job raises both latency and Buzz for marginal gain on early iterations. Quantity multiplies cost directly - quantity 3 is three billed renders, not one. Guidance and steps are the quality levers, and LTX's known weak spots (soft fine detail, prompt drift on complex inputs) respond better to prompt clarity than to cranking them. Your prompt, images and LoRAs leave your machine - it's a cloud job by design. And early preview: the whole pack's nodes and behavior can change without notice.