Nano Banana by Ace Data Cloud

Nano Banana for ComfyUI by Ace Data Cloud

Generate or edit images with Nano Banana, retrieve the same task ID, and load the completed image into ComfyUI's standard IMAGE type. These nodes call a hosted Ace Data Cloud API; no local model or GPU is needed. Generation and editing consume Ace Data Cloud Credits at the current service price.

简体中文 · Ace Data Cloud · Issues

| Node | Purpose | | --- | --- | | Nano Banana Generate Image | Submit one paid text-to-image task | | Nano Banana Edit Image | Submit one paid edit using public HTTPS reference image URLs | | Nano Banana Retrieve Task | Query or wait for the same task ID | | Nano Banana Retrieve Tasks | Read multiple existing task IDs without new generation | | Nano Banana Load Image / Load Images | Decode one result URL or a URL array into ComfyUI IMAGE |

The package offers all nine model IDs currently listed by the Ace Data Cloud Dify Nano Banana plugin:

nano-banana, nano-banana-2-lite, nano-banana-2, nano-banana-2.1, nano-banana-pro, nano-banana:official, nano-banana-2-lite:official, nano-banana-2:official, and nano-banana-pro:official.

Model availability, resolution limits, and pricing can differ. Check the current service documentation before a paid run. nano-banana and nano-banana-2-lite support only 1K; choose a model that supports 2K or 4K when requesting those sizes. A count of 1–4 is available; the Load Images node returns a batch when the output images have matching dimensions.

Current version and install source

This source version is 0.1.2. Check GitHub Releases for its versioned ZIP. The Registry section below retains the last verified Active 0.1.0 on 2026-10-19; that older ZIP does not contain this release's updated guide, examples or fixes. To use this version, install its GitHub Release ZIP, or select the exact 0.1.2 Registry release once it becomes Active. Do not substitute the initial nightly selection for a verified stable release.

To install the GitHub ZIP, stop ComfyUI and back up any existing node directory. Extract the ZIP contents into ComfyUI/custom_nodes/nano-banana/ without adding an extra nested directory, run python -m pip install -r requirements.txt there, then restart and refresh. Keep the API Token in the actual launch-process environment. A GitHub ZIP install is separate from a verified Registry/Manager stable installation.

Install

Open the official Nano Banana by Ace Data Cloud Registry page, or open Manager → Custom Nodes Manager and search Nano Banana. Use Ace Data Cloud as a fallback to find this publisher’s packages. Select Nano Banana by Ace Data Cloud, confirm publisher acedatacloud and source repository AceDataCloud/NanoBananaComfyUI, then click Install. In Select Version, explicitly select the approved Registry stable release 0.1.0. The normal dialog may initially highlight nightly; do not accept it for this stable-release tutorial. Keep dependency installation enabled. After Restart Required, restart ComfyUI and refresh the browser. Registry ID: nano-banana.

These are real 2026-10-16, ComfyUI 0.41.0 Linux screenshots from the publisher-brand installation test, not a new paid request or a screenshot of service-only search. See image sources and limits.

This Registry stable installation was verified on 2026-10-16 with original ComfyUI 0.41.0 on Linux CPU. That test did not submit a generation or chat request. Newer host versions and each paid action require their own checks; model availability and price remain subject to the current service rules.

For optional local source testing:

cd /path/to/ComfyUI/custom_nodes
git clone https://github.com/AceDataCloud/NanoBananaComfyUI.git
cd NanoBananaComfyUI
python -m pip install -r requirements.txt

Restart ComfyUI and search the node picker for Nano Banana. The package uses ComfyUI's existing PyTorch, NumPy, and Pillow installation for image decoding.

Set your Ace Data Cloud key

Copy or create your own API Token

The Applications screen is shared across Ace integrations. 1 copies the existing token; 2 opens Manage Keys, then Create. A new account signs in/registers and initializes General application first. Verify this service's access, current price, balance and any expiry/Allowed APIs restrictions.

These masked 2026-10-08 Ace Console screenshots are reused from SunoDify. The example credential name in the second image is Dify tutorial; name yours ComfyUI. No credential was created by this guide update. Paste only the token into the environment of the process that launches ComfyUI (ACEDATACLOUD_API_KEY), then restart. There is no ComfyUI API Key textbox in these nodes; do not paste it into prompt/JSON widgets or the workflow file. An unrelated terminal export does not configure an already-running Desktop process.

1. Sign in to Ace Data Cloud → Applications. If your account has no General application, choose Initialize General Application.
2. Open General application and check access to the service used by this package. A service-specific application key is limited to that service. Copy an existing API key, or choose Manage Keys → Create for a separate key named ComfyUI.
3. Copy only the token string, without Bearer or quotation marks. A platform management token is not the API key these nodes use. If you enable Allowed APIs, include every submission and task-query path used by your workflow. Check current pricing and available Credits before a paid action.

Use an application key with Nano Banana access. Set it in the ComfyUI process environment and restart:

export ACEDATACLOUD_API_KEY='paste-your-key-here'
python main.py --cpu

ACEDATACLOUD_API_TOKEN is also accepted. Keep the key out of workflows, screenshots, issues, and the repository. If you restrict API paths, allow /nano-banana/images and /nano-banana/tasks.

Generate and retrieve

Import 01-submit-nano-banana.json with File → Open, choose a model and price, then run it once. Copy the task ID from Preview as Text. A submitted task is not a completed image. If submission times out without a task ID, check Ace request history before paying for another generation.

Import 02-retrieve-and-preview.json, replace the placeholder task ID, and run it. It connects Retrieve Task → Load Images → Preview Image. Retrieve waits up to 240 seconds for the same ID. If still pending, rerun only the retrieve workflow. A finished task supplies image URLs; Load Images decodes them into a normal ComfyUI image batch.

For editing, import 03-edit-nano-banana.json and replace its placeholder source URL with a public HTTPS reference image. You can also enter a JSON array or one URL per line. It uses the same /nano-banana/images endpoint with action: edit. Reuse the retrieval workflow for the returned task ID. Reference images are submitted to Ace Data Cloud for processing.

Check usage and troubleshoot

Check the submitted task or trace ID in the Ace Data Cloud Console. Read-only task queries and media downloads should not create another generation.

| Symptom | Action | | --- | --- | | Key missing | Put ACEDATACLOUD_API_KEY in the actual ComfyUI process environment, then restart. | | HTTP 401/403 | Check key validity, service access, path restrictions, and content policy. | | HTTP 429 | Wait for the service limit; keep any accepted task ID. | | Pending | Query the same task ID again, without rerunning Generate or Edit. | | Submission timeout or HTTP 5xx | Inspect request history before resubmitting; the paid task may have been accepted. | | Image download error | Confirm the task succeeded and its Ace media URL is available. |

The nodes send prompts and reference URLs to api.acedata.cloud; completed images are downloaded from Ace CDN. See privacy notes. Only submit content you have permission to process.

Development

python -m unittest discover -s tests -v
python -m py_compile __init__.py client.py nodes.py

pyproject.toml declares the short Registry ID nano-banana, publisher acedatacloud, and a 400×400 proportional resize of the canonical Ace logo.

Task lookup nodes refresh the same task on every queued execution rather than reusing a cached pending result. Paid submissions retain stable caching when inputs are unchanged, so refreshing a query does not resubmit generation. Disable auto-queue/auto-run before importing or running these workflows, and explicitly submit once.