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

This is the "Channel Edition" of Grok's text-to-image node in BizyAirPlus. It sends your prompt to SiliconFlow's BizyAir channel, xAI's Grok Imagine model does the drawing on their servers, and ComfyUI gets back a normal image. No GPU, no xAI login, no model download - just a prompt and a per-generation bill.

Grok's image model is the one that powers image generation inside X, and the community reading on it has always been "least restrictive of the big closed models, with a taste for photorealistic people shots." If you want that flavor without signing up for xAI's own API, the channel edition exists precisely so one SiliconFlow key reaches a pile of these services. You're trading your usual local samplers for a hosted model, which makes sense when you want that model's look rather than something you could run yourself.

How it works

Under the hood this is a thin API wrapper, not a diffusion pipeline. The pack's BizyTRD nodes all work the same way: you flip the BizyAirPlus button in the action bar to ON, queue your workflow, and the node packages the prompt, fires it at the BizyAir channel, and turns the response into an IMAGE tensor when it lands. The local machine never runs Grok - which is exactly the point when your GPU is a potato or your VRAM is already spoken for.

Two optional inputs are worth knowing about. api_config lets you override base_url and api_key for a single request - handy if you're pointing at a different channel endpoint or a project-scoped key. skip_error, when enabled, won't kill your whole workflow on a failed request; the node outputs a placeholder of the expected type instead. In a big automation run that's often the difference between one bad batch and the entire queue dying.

Inputs and outputs that matter

• prompt (required, multiline) - the only thing you actually have to fill in.
• aspect_ratio (optional, default 16:9) - a combo box, so pick from the list rather than fighting with latent sizes.
• Outputs: image (IMAGE, wire into a save/preview node), plus request_id and urls (STRINGs - your reference for support and the hosted file if you'd rather grab it directly).

Installing BizyAirPlus

This node ships in the BizyAirPlus pack. Easiest route is ComfyUI Manager: search "BizyAirPlus" and install. Manual install works too:

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

Restart ComfyUI, then create an account at bizyair.ai, grab an API key, and flip the BizyAirPlus button ON. The key goes in the prompt that appears, or via Settings > BizyAirPlus > API Key, or the BIZYAIR_API_KEY environment variable. There are no model files to download - the pack pulls in bizyair-cloudberry and bizytrd as Python dependencies, nothing heavy.

Common issues

The two things that bite people here are the same two things that bite every BizyAirPlus node. First, dependencies installed into the wrong Python environment - check with python -m pip show bizyair-cloudberry; if the BizyAirPlus button never appears, that's your symptom. Second, auth and network: a bad key or a blocked connection produces a cloud-execution failure, and the detailed error lives in the ComfyUI console. Also remember these are paid, credit-based generations - a runaway batch loop is a real way to burn through a balance. Keep skip_error in mind and you'll be fine.