Symbiotica Wavespeed Nano Banana 2 Edit Fast
The quick iteration path for Nano Banana 2 edits
- images
- image
- task_id
- result_url
Last of the four Nano Banana 2 wrappers, and the least interesting on paper - but it's the one that maps to a real workflow. It's the Edit node's fast lane: same Google Gemini 3.1 Flash image model over Wavespeed, same up-to-14 reference images, fewer knobs, and a shorter default timeout because the fast endpoint is supposed to get back to you sooner. Use it when you're iterating on an edit - trying several prompts against the same reference sheet - and use the full Edit node when the result matters.
The mechanism is identical to the other three: images and prompt are posted to google/nano-banana-2/edit-fast on Wavespeed's API, the node polls until the task completes, and the first output image comes back as a ComfyUI IMAGE.
What's different from the full Edit node
resolutiononly offers2k(default) and4k.- No
enable_image_search- web-search grounding only (enable_web_search, off by default). timeoutdefaults to 180s instead of 300.
Everything else matches the Edit node: images (up to 14; bigger batches truncate with a console note), prompt (required, non-empty), aspect_ratio, output_format (png/jpeg), optional api_key. Outputs are the usual trio - image, task_id, result_url.
Install
One node of the Symbiotica pack:
cd ComfyUI/custom_nodes
git clone https://github.com/symbiotica-ai/comfyui-nodes.git symbiotica
pip install -r symbiotica/requirements.txt
Restart ComfyUI, or ComfyUI Manager → search Symbiotica → install. Needs a Wavespeed key - Settings → Symbiotica, or export WAVESPEED_API_KEY=.... No model downloads for this node.
Common issues
- Timeouts at 180s - fast endpoint or not, 4K edits with web search can run past the default. Bump
timeout. - Silent 14-image truncation - feed a big batch and only the first 14 frames go. For template-pipeline work, pass the sheets, not the whole folder.
- Refused edits - Google's filter applies at the model. Nothing the node can do about that; it's the cost of a closed model.
Four variants, one model, and a sensible division of labor: full node for the keepers, Fast for drafts. The EditFast is the draft lane for edits, and that's all it needs to be.
Inputs (8)
| Name | Type | Default | Description |
|---|---|---|---|
| images | IMAGE | — | |
| prompt | STRING | — | |
| aspect_ratio | COMBO | auto | 15 options: auto, 1:1, 3:2, 2:3, 3:4, 4:3, +9 |
| resolution | COMBO | 2k | 2 options: 2k, 4k |
| enable_web_search | BOOLEAN | false | — |
| output_format | COMBO | png | 2 options: png, jpeg |
| api_keyopt | STRING | — | |
| timeoutopt | INT | 18030–1800 | — |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| image | IMAGE | — |
| task_id | STRING | — |
| result_url | STRING | — |