WaveSpeedAI Google Nano Banana Pro T2I Multi

The text-to-image half of the Nano Banana Pro family, with the same twist as its edit sibling: it hands you two images per run, not one. From the razvanmatei-sf/razv-wavespeed pack, this node calls Google's Nano Banana Pro image model through the WaveSpeed AI API, and it's the model you'd pick when you care about rendering text, brand logos, and clean composition - the stuff Gemini's image models made their name on.

Why two images? The fixed num_images = 2 is the API's way of giving you a same-prompt pair, like a low-rent A/B test. You describe once, you get two takes, you keep the better one. It's billable as two generations, so the "one click, two results" convenience isn't free - but if you were going to roll the dice twice anyway, this saves you a queue.

How it works

Same skeleton as every node here: WaveSpeedAI Client → prompt → POST to the Nano Banana Pro endpoint → results converted to a batched IMAGE tensor on output_images. The aspect_ratio menu is short - 3:2, 2:3, 3:4, 4:3, defaulting to 3:2 - so landscape and square aren't on offer; plan your framing accordingly.

output_format (png/jpeg) carries the transparency rule from the rest of the family: png if you want alpha, jpeg if you don't. enable_sync_mode defaults on and blocks until both images land.

The inputs that matter

• prompt - the description. Nano Banana models reward concrete visual language; "a product shot, teal background, soft studio light" beats "nice product."
• aspect_ratio - four choices, no square. Pick the one that matches where the image will live.
• output_format - png for transparency, jpeg for small files.

Install

Shared pack:

cd ComfyUI/custom_nodes
git clone https://github.com/razvanmatei-sf/razv-wavespeed
# restart ComfyUI

Or ComfyUI Manager → search "razv-wavespeed" → Install. Add a WaveSpeed AI key through the WaveSpeedAI Client node, config.ini (copy from config.ini.tmp, [API]), or a WAVESPEED_API_KEY env var.

Common issues

• Two images = two bills. The node is upfront about the fixed count, but the cost is easy to miss if you're scanning. Budget for doubles.
• No square. If you need 1:1, this isn't the node - the plain Nano Banana text-to-image node has no aspect-ratio lock and effectively lets you pick up where this one can't.
• Paid and queue-backed, like everything in the pack. Expect occasional waits during peak hours.
• Batch output handling. output_images is a batch tensor; a single-image preview node will only show the first frame. Use batch-aware nodes to see both.