WaveSpeedAI Google Nano Banana Pro T2I Ultra

Here's the node with the honesty problem: its resolution dropdown literally says 4k ($0.15), 8k ($0.18). No other node in this pack shows you the price tag in the tooltip, and that's because this is the most expensive single image this pack can make. Nano Banana Pro Ultra is Google's top text-to-image tier, delivered at up to 8K through the WaveSpeed AI API - from the razvanmatei-sf/razv-wavespeed pack, so yes, it's another billable cloud call.

The appeal is straightforward: if you need a hero image at print or large-display resolution - and the model's famous text rendering to go with it - this is one of the few ways to get there without a workstation full of GPUs. The 8K number is the headline and the budget item; treat it like renting a nice dinner per image.

How it works

Standard pack wiring: WaveSpeedAI Client → prompt → POST to the Nano Banana Pro Ultra endpoint → the result comes back as an IMAGE tensor on output_image. The dials:

• resolution - 4k or 8k, and the tooltip prices both. 8K at ~$0.18 is a bargain compared to what the hardware would cost, but it's still $0.18 a click.
• aspect_ratio - the full ten: 1:1, 3:2, 2:3, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, and 21:9. The widest menu in the pack, and 21:9 is where Ultra earns its keep (ultrawide wallpapers, broadcast frames).
• output_format - png/jpeg, transparency rule applies.

The inputs that matter

• prompt - description. At this resolution, vague prompts get expensive fast; write for the detail level you're paying for.
• resolution - the cost dial. Start at 4k; most web uses don't need 8K.
• aspect_ratio - the widest choice set in the pack. Pick deliberately.

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 via the WaveSpeedAI Client node, config.ini (copy from config.ini.tmp, [API]), or a WAVESPEED_API_KEY env var.

Common issues

• The price is real and it's in the UI. The tooltip isn't flavor text - 8K costs ~$0.18 a pop plus queue time. If you're batch-generating, this is the fastest way to empty a small API credit balance.
• Big images, big latencies. Ultra generations are slow and the queue is real. Sync mode (default on) will hold your graph; consider async for anything exploratory.
• Don't upscale to save money. If you just want a bigger file, a 4K output through a local upscaler is cheaper than the 8K tier.
• Transparency: png for alpha, jpeg otherwise - the one rule that never changes in this pack.