Zhenzhen Nano Banana 2 Edit Fal

The pack gives you two doors into Nano Banana 2, and Comfly_nano_banana_2_fal ("Zhenzhen Nano Banana 2 Edit Fal") is the FAL one. The difference: where the Google-class nodes talk to the author's direct Gemini proxy, this one runs through the FAL-style queue - submit, poll, download - and it picks up two things the direct path doesn't emphasize: num_images (1 to 4 per call, so you can batch candidates in one submission) and a resolution switch up to 4K. If your workflow is "give me four candidates, pick the keeper, upscale," this node is built for that loop.

It's the FAL-class variant of the same model family the Comfly_nano_banana2_edit node covers. Input-wise it's broadly similar - prompt, up to eight reference images, aspect ratio, output format - but with FAL-specific extras. The FAL family takes the key in an api_key field directly, and the billing is pre-deduct: the site holds credits up front and settles after the run.

The inputs that matter

• prompt - required, the instruction.
• image1–image8 - reference images for edit jobs.
• num_images - 1 to 4. This is the batch button, and the price scales with it.
• aspect_ratio - auto plus the standard set.
• resolution - 1K (default), 2K, or 4K.
• output_format - png/jpeg/webp.
• safety_tolerance - 1 to 6; how much filtering you tolerate. The default 4 is the middle ground.
• system_prompt - an optional hidden-instruction field to steer the model's persona/style. Handy and underused.
• image_way - image_url or base64 for how reference images are sent.
• seed - a real FAL parameter here (0–65535, 0 = random), not just cache control.
• poll_interval / max_poll_attempts - note the defaults: 3s × 1200 = a one-hour timeout ceiling, which tells you the author expects slow 4K runs.

Outputs

• images - the result(s) as IMAGE.
• image_urls - hosted URLs as a string.
• response - the raw payload.

Installing it

cd ComfyUI/custom_nodes
git clone https://github.com/T8mars/Comfyui-zhenzhen
# restart ComfyUI

Or ComfyUI Manager → "Comfyui-zhenzhen", then a key from the author's site into api_key.

The honest part

The one-hour poll ceiling and the num_images batching tell you what this node is really for: hands-off batch generation where you can start a run and walk away. That's the FAL queue's strength. The weakness is the same one everywhere in this pack - a reseller proxy means your images and prompt go through a third party, and Google's congestion still bites (the 4K white-image problem lives on the direct path too; if the FAL route hits it, same fix - premium group on the site). Pick this node when you want batching and queue semantics; pick the direct edit node when you want synchronous, single-shot control.