Nano Banana Image Generation Editing

---
name: nano-banana-image-generation-editing
description: "Nano Banana 2 image generation and advanced editing on PoYo / poyo.ai via `https://api.poyo.ai/api/generate/submit`; use for `nano-banana-2-new`, `nano-banana-2-new-edit`, multi-reference workflows, text-to-image, image-to-image, and 1K/2K/4K output."
category: "Media"
author: community
version: "1.0.2"
icon: image
---

# PoYo Nano Banana 2 Image Generation and Advanced Editing

Use this skill for Nano Banana 2 jobs on PoYo. It fits multi-reference generation, advanced editing, and higher-resolution output; read `references/frontend-notes.md` when product-side defaults matter.

## Use When

- The user explicitly asks for `Nano Banana 2`, `nano-banana-2-new`, or `nano-banana-2-new-edit`.
- The task depends on multiple reference images, image composition, or advanced edits.
- The workflow needs `1K`, `2K`, or `4K` output, or may use `google_search`.

## Model Selection

- `nano-banana-2-new`: text-to-image and image-to-image entry point.
- `nano-banana-2-new-edit`: use whenever the user explicitly wants to edit supplied images.

## Key Inputs

- `prompt` is required and limited to 1000 chars.
- `image_urls` supports up to 14 reference images.
- `size` supports `1:1`, `2:3`, `3:2`, `3:4`, `4:3`, `4:5`, `5:4`, `9:16`, `16:9`, `21:9`.
- `resolution` supports `1K`, `2K`, `4K` and defaults to `1K`.
- `google_search` is optional and only useful for real-world grounding.

## Execution

- Read `references/api.md` for payload fields, examples, and polling notes.
- Read `references/frontend-notes.md` when you need product-side defaults or want to mirror the current desktop frontend behavior.
- Use `scripts/submit_nano_banana_2.sh` to submit a task from the shell when direct API execution is appropriate.
- If the user needs a raw curl example, adapt one from `references/api.md`.
- After submission, report the `task_id` clearly so follow-up polling is easy.

## Output expectations

When helping with a Nano Banana 2 task, include:
- chosen model
- final payload or summarized parameters
- whether reference images are used
- returned `task_id` if the request was actually submitted
- next step: poll status or wait for webhook