Google Nano Banana Pro Text To Image (Official Edition)

Nano Banana Pro is Google's flagship-grade image model - closed weights, no local version, and one of the strongest pure text-to-image engines you can reach from ComfyUI without a cloud GPU. The Official Edition node is where BizyAir exposes the whole vendor parameter surface: the usual prompt/resolution/aspect controls plus seed, temperature, top_p, and max_tokens. If you've ever wanted to actually tune a hosted image model instead of just pointing it at a prompt, this is the node for you. If you'd rather not think about any of that, the Channel Edition will happily keep your life simple.

What "official" buys you

Every model in BizyAir ships two ways. The _Base_/Channel classes call BizyAir's own serving channel with a trimmed-down set of knobs; the _Official_ classes call the vendor's official API endpoint - here nano-banana-pro-official/text-to-image - and mirror its real parameter surface. Same key, same gateway, more control. For text-to-image, that mostly means the sampling dials: temperature (default 0.95), top_p (default 0.95), and max_tokens (default 1 - that's the tooltip's stated default, odd as it looks; raise it if generation truncates). The defaults are sane for most work; the dials exist because sometimes you want a model that's less creative or more reproducible.

How it works

Cloud API node, full stop. The prompt goes to api.bizyair.ai, the job runs remotely on the nano-banana-pro-official endpoint, and the result is downloaded back as a native IMAGE tensor. No GPU work, no model files, and a metered bill on your BizyAir credits for each job.

Inputs that matter

• prompt - the entire creative brief; Pro handles long, detailed descriptions well.
• resolution / aspect_ratio - dropdowns, 2K and 16:9 by default.
• seed - integer, default 0. Set it for reproducible output.
• temperature / top_p - sampling creativity controls. Lower temperature for tighter, more literal results; leave near 1.0 for variety.
• max_tokens - output length cap; the default 1 is a quirk, raise it if your generations come back short.
• api_config / skip_error - the standard pair: per-request key/base override, and the fail-safe that returns placeholders instead of crashing the workflow.

Outputs: image (to preview/save), request_id, urls.

Install and get going

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, register at bizyair.ai, set the key in Settings > BizyAirPlus > API Key (or BIZYAIR_API_KEY), and flip the BizyAirPlus action-bar toggle to ON. Every BizyTRD node is gated on that switch.

Troubleshooting

The standard BizyAir failure stack: missing key, cloud toggle off, or the runtime installed into the wrong Python (python -m pip show bizyair-cloudberry to check). The console holds the real error. One tuning note specific to this node: if output quality seems "off," it's usually temperature sitting too high for your use case rather than the model being weak - drop it toward 0.7 and compare before you start spending credits on retries.