grok / v1.0 / text-to-video

xAI's Grok family has always been the interesting closed corner of the ecosystem - the image side is an autoregressive mixture-of-experts that early on looked suspiciously Flux-based, and the least-restrictive moderation around. This node is its text-to-video surface, v1.0: type a prompt, get a clip, pay in Buzz. The model is closed - there are no Grok video weights to run locally, full stop - so if you want it inside a ComfyUI graph, an API wrapper like this is the only door.

It's part of civitai-comfy-nodes, Civitai's official pack of ~160 nodes generated straight from its Orchestration API spec. Mechanism-wise it's the same as every node here: the node builds a videoGen job tagged grok / v1.0 / text-to-video, submits it to Civitai's cloud, long-polls until it succeeds, and returns a native VIDEO tensor. Nothing runs on your machine.

What you set

One required input, prompt (multiline). Everything else is optional:

• duration - 1 to 15 seconds, default 6. Video is billed per second, so this is your cost lever.
• resolution - 480p or 720p. No 1080p on v1.0; that arrives in v1.5.
• aspect_ratio - no auto here, you pick: 16:9 default, then 4:3, 3:2, 1:1, 2:3, 3:4, 9:16.

Outputs are the standard pack set: video (VIDEO), draft_cache_audio (AUDIO), draft_cache_video (VIDEO), workflow_id and raw_json (STRING). The node stamps its workflow id and cost onto itself after each run.

Install and auth

Official pack, Comfy Registry, one dependency:

# 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, then authenticate: a Civitai Auth node, the CIVITAI_API_TOKEN env var (best for headless), or stored OAuth from the Civitai sidebar.

Common issues

The pack-level ones apply: Buzz spend - every second of output is billed and retries compound it; your prompt leaves your machine to a server with its own logging; and the whole pack is early preview with nodes that "may change without notice." One Grok-specific note: v1.0's 720p cap and 15-second ceiling make it the budget tier of the family. If you want 1080p, image-conditioning, or the referenceToVideo mode with taggable reference images, that's the v1.5 nodes - same menu, clearly named. For pure prompt-to-clip with the fewest dials, this one is the simplest Grok node in the pack, which is a real virtue when you're iterating on phrasing.