EvoLink Grok Imagine Text-to-Video (Official)

EvoLink_GrokVideoT2V is xAI's text-to-video, and it's the node to reach for when you want a moving shot straight from a prompt and you don't need 4K polish. Grok's video tier is a bit rough around the edges compared to the top-end closed models, but it has two things going for it: it's cheap, and it's got the loosest-feeling guardrails of the bunch, exposed here as a mode setting with normal, fun, and spicy. In a pack full of serious video models, this one has a personality.

Keep expectations calibrated, though. This node tops out at 720p, so it's for concept work, storyboards, and quick drafts - not your hero render.

The inputs that matter

• prompt (required) - shot and motion description, Chinese-friendly.
• mode - normal (default), fun, or spicy. fun tends toward playful, spicy toward bolder. If you're chasing subject matter the bigger platforms refuse, this is the dial people talk about.
• duration - 6–30 seconds, default 6, billed by duration.
• quality - 480p (default) or 720p. There's no higher tier here.
• aspect_ratio - 16:9 (default), 9:16, 1:1, 3:2, 2:3.

Plus the pack-standard api_key (fill once, saved locally, leave blank after) and timeout_seconds (default 1200).

How it works

The node submits your prompt to POST /v1/videos/generations and polls the task endpoint until the clip completes, with a progress bar on the node. The finished video is downloaded to your machine as a VIDEO object - connect the video output to a native Save Video node to keep it. The other four outputs (task_id, status_info, result_urls, response_json) are the pack-wide standard, and the raw result links expire after 24 hours. Failed or content-blocked jobs don't charge, and rerunning unchanged params reuses ComfyUI's cache.

Because the output is a VIDEO type, this needs a recent ComfyUI with comfy_api's VideoFromFile; on an older install you'll get an explicit error about missing VIDEO support.

Installing

No GPU, no model files, only requests:

cd ComfyUI/custom_nodes
git clone https://github.com/deeplearning-goethe/comfyui-evolink.git

then fully restart ComfyUI. Or install via ComfyUI Manager by searching EvoLink. If the node's missing, check for a nested folder (the pack's __init__.py must sit directly in custom_nodes/comfyui-evolink/).

Troubleshooting

401 bad key, 402 out of credits (check first with the EvoLink Credits node), 403 no model access, 429 rate limited. VPN users should whitelist api.evolink.ai and files-api.evolink.ai. And a realistic note on moderation: spicy is a dial, not a bypass - xAI's moderation still runs on their servers, so content that's refused stays refused. Tone it down and retry; refused jobs aren't charged.