SF WaveSpeed Nano Banana Edit
Edit an image by describing the change — the closed Gemini editor, by API
- client
- image
Nano Banana is Google's Gemini-based image model, and the community's verdict landed fast: for editing - "put her in a car, same pose" - it was, in the words of a thread that ran up a few hundred upvotes, probably the best edit model out there when it launched. You can't download it; it's closed and runs on Google's side. SF WaveSpeed Nano Banana Edit serves it over the WaveSpeed API, so you describe the edit in a sentence, point at an image, and get back a real IMAGE tensor - no Gemini app, no Gemini key, just your WaveSpeed key.
How it works
The node takes a prompt and an input image URL, sends them to WaveSpeed's google/nano-banana/edit endpoint, and returns the edited image. Two details make this node friendlier than its siblings. First, output_format defaults to PNG, and the tooltip explains why: PNG preserves transparency, and Nano Banana is unusually good at honoring "keep the alpha" when you're working with cutouts and logos. Second, there's an optional additional_images field - URLs, one per line, up to 9 more - because Nano Banana accepts multiple reference images. That's the feature that makes it a real editing tool rather than a glorified filter: feed it a reference for style or identity alongside the main edit target.
The inputs that matter
- client - from SF WaveSpeed Client, wired first.
- prompt - the edit instruction, in plain language. This model rewards describing the change precisely: "Replace the cloudy sky with a clear sunset" beats "make it better."
- image_url - the input image as a public URL.
- output_format - PNG (default) for transparency work, JPEG when you don't care.
- enable_sync_mode - default true; leave it.
Output: image (IMAGE), which plugs straight into the rest of your graph.
Installing it
In the SF ComfyUI Nodes pack from Stillfront. ComfyUI Manager → "SF ComfyUI Nodes" → Install → restart, or:
cd ComfyUI/custom_nodes
git clone https://github.com/Stillfront/comfyui-sf-nodes.git
cd comfyui-sf-nodes
pip install -r requirements.txt
Plus a WaveSpeed API key through SF WaveSpeed Client.
Common issues
The URL requirement is the recurring theme - your image_url and any additional_images must be publicly fetchable or the request fails before generation starts. The subtler gotcha is prompt quality: Nano Banana is instruction-following in a way diffusion models aren't, so vague prompts produce confidently wrong results. The famous comparison thread showed that a muddled prompt made even the best model stumble while a precise one nailed it. And remember it's Google under the hood - the model has safety filters baked in, so don't expect it to do anything your favorite open model would politely decline.
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| client | WAVESPEED_AI_API_CLIENT | — | |
| prompt | STRING | The positive prompt for image generation | |
| image_url | STRING | URL of input image for editing (or connect from Upload Image node) | |
| output_format | COMBO | png | Output format - use PNG for transparency support |
| enable_sync_mode | BOOLEAN | true | Wait for image generation to complete before returning |
| additional_imagesopt | STRING | Additional image URLs (one per line, max 9 additional) |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| image | IMAGE | — |