NS WaveSpeed Nano Banana Edit

Nano Banana is Google's consumer image model - Gemini 2.5 Flash Image wearing a rebranded name - and "edit" is what it does best: give it an image and a sentence, get back the image changed to match. This node is WaveSpeed's hosted version, so the whole exchange happens over the API and the result lands in your graph as a normal IMAGE tensor.

It's part of the Symbiotica pack's neuralsins/WaveSpeed family. The appeal is the same one the KB's closed-source doc flags: Nano Banana has no open weights, so "inside ComfyUI" only happens through an API. This node is one of the cheaper ways to get it there - no Comfy account, no prepaid credit storefront, just your WaveSpeed key and per-call metering.

How it works

You give it an image_url and a prompt, it POSTs to WaveSpeed's /api/v3/google/nano-banana/edit, and the edited image comes back through image (IMAGE). Editing here is instruction-following, not mask-based - you describe the change and the model re-renders. That's powerful for "change the sky, keep the subject" jobs and weaker for "touch exactly these pixels and nothing else" work, which is still a mask-and-inpaint job locally.

The inputs you set:

• prompt - the edit, in plain language. The model understands English and is genuinely good at following a detailed sentence.
• image_url - the source, as a URL string. Same hosted-URL rule as every node in this family.
• additional_images (optional) - extra reference URLs, one per line, up to 9 more. This is how you do style or subject references - show it the look you want along with the image to edit.
• output_format - png (default, and the tooltip notes it's for transparency support) or jpeg.

enable_sync_mode is on by default, so the node waits and hands back the finished tensor rather than a job id.

Installing

Install the whole pack once:

cd ComfyUI/custom_nodes
git clone https://github.com/symbiotica-ai/comfyui-nodes.git symbiotica
pip install -r symbiotica/requirements.txt

Restart, or search Symbiotica in ComfyUI Manager. Add an NS WaveSpeed Client node with your key (Settings → Symbiotica, or WAVESPEED_API_KEY), and feed its client output into the socket.

Common issues

• "Connect from Upload Image node" - image_url wants a reachable URL, not a local tensor. This is the #1 stumbling block for the whole family.
• The edit drifts other details. That's the nature of whole-image instruction editing - Nano Banana will happily change things you didn't mention. Keep prompts tight and specific.
• Transparency vanishes. If you picked jpeg, that's on you - jpeg can't carry alpha. Use png when the result feeds another composite.
• Every run is a paid call. The meter runs per image, so mute the branch when iterating on the prompt wording.

The honest take: this is the node you reach for when you want Nano Banana's editing in the middle of a bigger workflow rather than as a one-off website visit. It's not a replacement for local editing - it's the closed model your local stack can't offer.