Google Nano Banana Pro Edit (Official Edition)

Instruction-based editing is the Nano Banana line's signature move - the thing the community spent all of 2025 crowing about in comparison threads. Nano Banana Pro does it with more polish and patience than the base model, and this node wraps it as the Official Edition: same BizyAir key, vendor API endpoint, and the full complement of controls including seed, temperature, and top_p. It's the node to reach for when you want high-fidelity, tunable image editing in ComfyUI and you're comfortable paying per job on a cloud API.

What this node is doing

Pure cloud execution. Your reference images upload to api.bizyair.ai, the request goes to the nano-banana-pro-official/image-to-image endpoint, and the edited result is downloaded back into ComfyUI as a native IMAGE tensor. Nothing runs on your GPU, nothing lives in your models/ folder. The flip side, stated plainly: your source images pass through SiliconFlow's servers and each job draws from your BizyAir credits.

Inputs that matter

• images - the edit source, an autogrowing input. One image for a simple edit; several for identity/style juggling ("keep her face, his jacket, the room, and swap the car").
• prompt - the instruction. The reliable pattern is what stays + what changes, in plain language. "Keep the pose and lighting, change the setting to a rainy street at night" will beat "rainy street night" every time.
• resolution - defaults to 2K; aspect_ratio defaults to 16:9.
• seed - integer, default 0; set it once you land on a look you want to reproduce.
• temperature / top_p - sampling creativity dials (both default 0.95). For surgical edits, a slightly lower temperature keeps the model closer to your instruction.
• max_tokens - output length cap; if an edit comes back incomplete, raise it from the default.
• api_config / skip_error - the standard pair: per-request key/base override, and placeholder-on-error.

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

Install and first run

ComfyUI Manager → "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 ON - the node refuses to run while that's off.

Gotchas

Setup failures are almost always the same trio: missing key, toggle off, or deps in the wrong Python (python -m pip show bizyair-cloudberry to verify). On the editing side, the expensive failure is a lazy prompt - the model follows instructions literally, so "remove the watermark" leaves you with a different watermark if you didn't say exactly which one. Since retries cost credits, spell out the constraints before you queue.