Google Nano Banana 2 Edit (Official Edition)
Nano Banana 2's superpower is editing — here's the official I2I node
- images
- api_config
- image
- request_id
- urls
Here's the thing about Nano Banana: text-to-image is almost beside the point. What made the community lose its mind over Google's Gemini image model - the threads literally titled "Nano banana killed it" in the Qwen-Image-Edit comparison wars - was instruction-based editing. "Put her in the car with the same pose." "Change this to rain." That's this node's whole job: feed it images, tell it what to change, get a coherent edit back. And it's the Official Edition, so you get the full Google-API parameter surface through BizyAir instead of the stripped-down channel version.
What you're actually running
This is a cloud API node. The images you feed it are uploaded to api.bizyair.ai, the request goes to the nano-banana-2-official/image-to-image endpoint, and the finished edit is downloaded back into ComfyUI as a native IMAGE tensor. Zero GPU work on your side, zero model files. The bill lands on your BizyAir credits, and the images themselves spend some time on SiliconFlow's servers - worth knowing before you feed it something private.
Inputs that matter
- images - the reference input, an autogrowing slot that accepts one or more images. For a single-edit job, one is plenty; Nano Banana genuinely handles multi-image references (identity, style, "use this outfit on this person").
- prompt - the edit instruction in plain language. Be specific about what stays the same; the model is best when you say "keep the pose and lighting, but change X".
- resolution - defaults to
2K. - aspect_ratio - defaults to
16:9. - seed - integer, default
0. Set it to lock in a result you like. - web_search - defaults to on, letting the model reach for real-world references mid-edit. Turn it off for strictly self-contained edits.
- api_config / skip_error - per-request base_url/key override, and the fail-safe that returns a placeholder image instead of crashing your workflow.
Outputs are image, request_id, and urls - wire image into a preview or save node and you're done; the other two are for support and direct download links.
Install and get going
Install the pack via 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, create an account at bizyair.ai, and set the API key under Settings > BizyAirPlus > API Key (or via the BIZYAIR_API_KEY env var). Then click the BizyAirPlus action-bar button ON - the node does nothing until that cloud toggle reads ON.
Gotchas
The classic "it's broken" moment is really "the key's missing or the toggle's off" - the workflow just sits there. Check the ComfyUI console for the actual error. One genuinely Nano-Banana-specific tip: for edits, describe the change as an instruction, not a caption, and explicitly name what must not change. Vague prompts get you a pretty but wrong image, and unlike a local sampler you're paying per attempt.
Inputs (8)
| Name | Type | Default | Description |
|---|---|---|---|
| images | COMFY_AUTOGROW_V3 | — | |
| prompt | STRING | Prompts | |
| resolution | COMBO | 2K | Resolution |
| aspect_ratioopt | COMBO | 16:9 | Aspect ratio |
| seedopt | INT | 0-1–2147483647 | Input range value: 0-2147483647 seed integer to control randomness of generated content. |
| web_searchopt | BOOLEAN | true | Internet search |
| api_configopt | BIZYAIR_OPENAPI_CONFIG | 单次请求覆盖base_url和api_key | |
| skip_erroropt | BOOLEAN | false | 开启后遇到错误不中断工作流,输出对应类型的错误占位符 |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| image | IMAGE | — |
| request_id | STRING | — |
| urls | STRING | — |