X AI Grok Imagine Video Image To Video (Channel Edition)

xAI's Grok Imagine video model, image-to-video, routed through SiliconFlow's BizyAir channel. Feed it one or more images and a prompt, and it returns a video clip - 6 to 30 seconds long - that turns the still into motion. This is the "Channel Edition", meaning the request goes through the BizyAir channel with your SiliconFlow key rather than xAI's own API.

Grok Imagine's video output sits in the photorealistic tier of hosted video models - the same category as Veo and Kling - and it's notably good at preserving the identity and setting of the input image. That's the real reason to use I2V instead of text-to-video here: the model hangs onto the character and scene you give it, which is where a lot of hosted T2V models fall apart. If you've already got a strong still from your local pipeline, this node turns it into a clip without you touching video hardware, video training, or temporal conditioning at all.

How it works

Another thin cloud wrapper in the BizyTRD family. With BizyAirPlus turned ON, the node uploads your images, sends the prompt and duration, and the channel runs Grok's video model. The result comes back as a VIDEO tensor you can wire straight into a video preview or save node. Nothing runs locally - which for video is a big deal, because video inference is exactly where a normal consumer GPU runs out of road.

duration here is an integer slider (6 to 30 seconds, default 10) - note that the official edition of this same model takes a fixed duration combo instead, so the channel version is actually the more flexible one for clip length. resolution (default 720p) and aspect_ratio (default 16:9) are combos. api_config overrides the endpoint and key per request; skip_error swaps a failure for a placeholder rather than aborting your workflow.

Inputs and outputs

• images (required) - your starting still(s).
• prompt (required, multiline) - "Prompt words for video generation," per the tooltip.
• duration (required, 6–30s, default 10) - how long the clip runs.
• resolution (required, default 720p) and aspect_ratio (required, default 16:9).
• Outputs: video (VIDEO), request_id (STRING), urls (STRING).

Installing BizyAirPlus

ComfyUI Manager: search "BizyAirPlus". Or:

cd /path/to/ComfyUI/custom_nodes
git clone https://github.com/siliconflow/BizyAirPlus.git
cd BizyAirPlus
python -m pip install -r requirements.txt

Restart, get an API key at bizyair.ai, and flip the BizyAirPlus button ON. No model downloads; dependencies are bizyair-cloudberry and bizytrd.

Common issues

Missing button → dependencies in the wrong Python environment; python -m pip show bizyair-cloudberry confirms. Cloud failure → key, network, console. The video-specific one to watch: longer clips cost more, and a duration of 30s on a looped batch is an excellent way to discover your credit balance the hard way. Start short, verify quality on one clip, then scale.