SF WaveSpeed Nano Banana Pro Edit Ultra
Nano Banana Pro Edit at 4K or 8K — the big-canvas tier
- client
- output_image
The regular Nano Banana Pro Edit tops out at 4K. SF WaveSpeed Nano Banana Pro Edit Ultra goes one further - 4K or 8K output - and it's the node you reach for when the edit is going to be a hero image, a print, or a canvas someone will zoom into. Same edit model, same instruction-following, bigger canvas. The tooltip does the math for you: 4k at $0.15, 8k at $0.18 per image. Yes, that's an 8K edit for eighteen cents. The model is closed and rentable per-call, which is the only reason this is even possible in a ComfyUI workflow.
How it works
Mechanically it's the same pipeline as NanoBananaProEdit - prompt plus up to six image URLs, bundled into an images array, sent to WaveSpeed's google/nano-banana-pro/edit-ultra endpoint - but the resolution dial has moved up a tier: 4k and 8k, with 8k as the default. It returns a single IMAGE tensor at whatever resolution you picked. The trade-off to keep in mind: bigger output means a heavier call, so unless the deliverable genuinely needs the resolution, the standard edit node at 1k/2k is a better everyday driver. Ultra is for the final answer, not the first draft.
The inputs that matter
- client - from SF WaveSpeed Client, wired first.
- prompt - the edit instruction; specificity matters more the larger the canvas, because the model has that much more room to get details wrong.
- image_1 … image_6 - public URLs,
image_1required, the rest optional references. - aspect_ratio - default 1:1.
- resolution -
4k($0.15) or8k($0.18). 8k is the default; drop to 4k for most work. - output_format - PNG (default) for transparency.
- enable_sync_mode - default true.
Output: output_image (IMAGE).
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
Input images still need to be public URLs - that requirement never goes away in this pack. The more Ultra-specific gotcha is resolution expectations: 8K is a big tensor, and downstream nodes (preview, save, any processing you chain after) need to handle it without surprise. If ComfyUI starts swapping or choking after this node, that's your sign to process at 4k and upscale at the very end instead. And as always with the Nano Banana line, Google's safety filtering is in the weights - at 8K, a refused prompt is just a bigger refusal. For clean commercial-grade edits at print resolution, this is genuinely the one you'd reach for.
Inputs (12)
| Name | Type | Default | Description |
|---|---|---|---|
| client | WAVESPEED_AI_API_CLIENT | — | |
| prompt | STRING | Text description of the edit to perform (e.g., 'Replace the cloudy sky with a clear sunset') | |
| image_1 | STRING | Primary input image URL (connect from Upload Image node) | |
| image_2 | STRING | Second input image URL (optional) | |
| image_3 | STRING | Third input image URL (optional) | |
| image_4 | STRING | Fourth input image URL (optional) | |
| image_5 | STRING | Fifth input image URL (optional) | |
| image_6 | STRING | Sixth input image URL (optional) | |
| aspect_ratio | COMBO | 1:1 | Aspect ratio of the output image |
| resolution | COMBO | 4k | Output resolution: 4k ($0.15), 8k ($0.18) |
| output_format | COMBO | png | Output format - use PNG for transparency support |
| enable_sync_mode | BOOLEAN | true | Wait for generation to complete before returning |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| output_image | IMAGE | — |