grok / v1.5 / imageToVideo

Grok v1.5 is where xAI's video line grew up, and this node is its image-to-video entry: one or more start images, a prompt, and a clip comes back at up to 1080p. If you were hitting the 720p ceiling on the v1.0 node, this is the upgrade - same family, same closed model, but the resolution cap is gone and the auto aspect ratio has been dropped for an explicit pick. Which honestly tells you something about how the two versions are tuned.

Like every node in civitai-comfy-nodes - Civitai's official ~160-node pack generated from its Orchestration API spec - nothing runs locally. The node submits a videoGen job tagged grok / v1.5 / imageToVideo (note the camelCase - v1.0 used image-to-video; the naming changed between versions), uploads your stills, long-polls, and returns a native VIDEO tensor. Billed in Buzz.

Inputs that matter

prompt is required; the stills come through images, an IMAGE socket that takes a list. The dials are minimal:

• duration - 1 to 15 seconds, default 6.
• resolution - 480p, 720p, or 1080p (the v1.5 addition).
• images - feed one or several stills. For a single strong start frame this behaves like the v1.0 node; for style-and-subject driving across multiple references, the v1.5 referenceToVideo node is the better tool.

Outputs are the pack standard: 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 every run.

Install and auth

The usual for this pack - official, Comfy Registry, one requirements file:

# 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 and authenticate via a Civitai Auth node, the CIVITAI_API_TOKEN env var, or sidebar OAuth.

Where people get burned

Resolution is the price dial. 1080p output costs more per second than 720p, and with a default 6-second clip that difference multiplies fast - draft your motion at 720p, promote to 1080p only for the keeper. Your images leave your machine to a third-party server; keep that in mind before wiring in anything sensitive. Early preview applies to the whole pack - the README says nodes and behavior "may change without notice." And the v1.0-vs-v1.5 confusion is real: both versions have image-to-video nodes, and the only visible difference is the version number and the naming style. If you want the 1080p option, make sure you're grabbing the v1.5 node; the menu also contains a v1.5 referenceToVideo whose images input works differently, so read the node name before you wire a batch of stills in expecting start-frame behavior.