🍌 Banana｜普通五视图并发（5输出）

The "🍌 Banana｜普通五视图并发（5输出）" node is the three-view node's bigger sibling, and the math is exactly what the name says: five prompts, five concurrent API calls, five IMAGE outputs, one run. It's for when front/side/back isn't enough coverage - a product listing that wants a full orbit of a garment or object.

This is a rare case where the extra views aren't filler. The five slots are view_1 through view_5, and their defaults walk a real 360-ish sequence: front, left 45°, side (90°), rear 45°, back. That's the coverage a fashion listing actually uses - the angled shots are the ones that show drape and thickness, which dead-on views flatten out. If you've ever built a product-photography workflow locally, this is the shot list you'd end up writing yourself, minus the hours of prompt iteration (for the API-vs-local tradeoffs - cost, data leaving the machine, the closed backend - the KB's external-api-nodes essay is the short version).

How it works

One node, five independent prompt boxes, five parallel requests to the Hrio backend with the same reference images, model, image_size, and aspect_ratio. Same engine as the three-view node, just multiplied. The reference slots (image_1…image_10) are reused across all five requests, so consistency between angles comes from the backend doing the same subject with the same references - you're not hand-aligning anything.

The generate_scope dropdown here has six options: generate all five, or regenerate only view 1 through 5 individually. That's the part to remember. The other four views come back from the node's last successful run - cached, not re-billed. On a metered API where each view is a separate paid call, regenerating just the one bad angle instead of the whole set is where this node pays for itself.

Inputs and outputs

• api_key - the vendor key; blank falls back to the pack's config.ini.
• view_1_prompt … view_5_prompt - your text per angle. The defaults are solid Chinese e-commerce phrasing; they encode "pure white background, single subject, no collage, no watermark," which is the style the backend is tuned for.
• model - Banano (default), Banano Pro, or Gemini 3.5 Flash - display names for backend endpoints.
• image_size (2K default) / aspect_ratio (Auto default) - shared across all five.

Outputs: view_1_image through view_5_image, a combined images batch, plus info (STRING) and mp4url (STRING, empty here - that's for the video nodes).

One technical wrinkle worth knowing: this node class went through several renames in the pack's history (the current one is HrioBananaNormalFiveViewConcurrentNode), and the loader still hunts for legacy class names so old workflows keep loading. If you see it in the menu under the five-view display name, you're on the current build.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/moki-flank/Hrio-Commerce

or search Hrio-Commerce in ComfyUI Manager and restart. Dependencies are light (requests, numpy, pillow; torch comes with ComfyUI). Nothing to download - it's an API pack, so the only "model file" you need is the vendor's API key, in the node or in config.ini's [banana] section.

Common issues

• Five retries stack up. There's no retry toggle on this node (that's the template node's feature), but a flaky backend will still make five concurrent requests feel slow. Check the info string if outputs go missing.
• Key failure = error image. No key, and the node hands you a red error picture as output rather than a clean exception.
• It's proprietary and phones home. Closed-source pack, credential-holding, network-calling by design. Inspect it before you run it, and know your product images go to the vendor's server.
• Stale nodes after updates. Delete old Banana nodes and hard-refresh (Ctrl+F5) if behavior goes weird after a pack update.

Want fewer angles with simpler prompts? The three-view node. Want the guided templates? The panel node.