EvoLink Grok Imagine Video 1.5 (Official)

Most of this pack's video nodes are stubbornly one-mode: this one does image-to-video, that one only text-to-video. EvoLink_GrokVideo15 is the exception, and it's a nice one. It wraps xAI's Grok Imagine Video 1.5, and the model figures out what you want from what you feed it:

• No image connected → text-to-video.
• One image connected → first-frame image-to-video (that image starts the clip).
• Two to seven images connected → reference-to-video, holding the subject consistent across the shot.

Same node, three behaviors, no mode switch to forget. It also reaches up to 1080p, which is more than Grok's older video tier offered, and durations from 1 to 15 seconds.

The inputs that matter

• prompt (required) - the shot description. In reference mode, describe what happens; the model keeps the subject from your images.
• image1, image2, image3 - the sockets that decide the mode. One image = first frame; two or more = reference mode (up to seven total if you add image_urls_text with one public URL per line).
• quality - 480p (default, cheapest), 720p, or 1080p.
• duration - 1–15 seconds, any integer, default 6. Billed by duration, so shorter really is cheaper.
• aspect_ratio - auto (default) plus 16:9, 9:16, 1:1, 3:2, 2:3. In image mode the frame largely follows the input image.

Then the pack-standard api_key and timeout_seconds (default 1200).

How it works

Same machinery as every node here: reference images go up through EvoLink's file channel (deleted after 72h, no third-party host), the task is submitted to POST /v1/videos/generations, and the pack polls until the clip finishes, downloading it as a VIDEO object. Connect the video output to a native Save Video node to keep it. The other outputs - task_id, status_info, result_urls (24h expiry), response_json - are the pack-wide standard. Because the output is a VIDEO type, you need a recent ComfyUI with comfy_api support; an old install errors out with a message about missing VIDEO output.

Installing

The whole pack is light - no GPU, no model downloads, only requests:

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

then fully restart ComfyUI. Or search EvoLink in ComfyUI Manager.

Troubleshooting

401 invalid key, 402 empty balance, 403 no model access, 429 rate limited. VPN users should whitelist api.evolink.ai and files-api.evolink.ai for direct connection - the pack's client code flags this explicitly. The mode behavior is the thing to keep an eye on: if you meant reference mode and connected only one image, you're in first-frame mode instead, and the subject consistency won't hold. And remember, an unconnected video output means no visible result - wire it to Save Video.