Skip to main content

​
Short Answer

This is a known issue with the Google Aistudio API. Under compute pressure, large pure-white regions in generated images tend to render with black spots, dirty patches, or blurry color blocks. It has been observed repeatedly since April 2026. Two fixes:
  • Option 1 (try this first): change the prompt from “white background” to a light tint (light gray, beige, soft blue) — no configuration change required.
  • Option 2: contact support to enable the VertexGemini group, then create a token bound to that group. The Vertex channel produces clean pure-white backgrounds.

​
Detailed Explanation

Under load, the Aistudio image pipeline has trouble compressing and reconstructing large uniform-color regions, especially pure white. Three artifacts are common:
  • Black spots / dirty patches: localized dark spots or irregular blotches in the image
  • Blurry color blocks: instead of pure white, you get a hazy grayish or yellowish area
  • Color cast: the white background drifts warm (beige) or cool (grayish blue) versus the prompt
This is not a model capability issue — it is a rendering defect in the Aistudio pipeline on uniform backgrounds. Gemini image models such as Nano Banana Pro and Nano Banana 2 are all affected when routed through Aistudio.

​
Resolution Steps

Replace “white background / pure white background / #FFFFFF” in the prompt with a light tint:
  • Light gray: light gray background / #F5F5F5
  • Beige: beige background / #F5F0E5
  • Soft blue: light blue background / #E8F0F8
  • Transparent: transparent background (if the model supports alpha output)
In practice, switching from pure white to light gray or beige almost completely eliminates black spots and dirty patches. The subject colors are not affected — only the background tint changes.

​
Option 2: Move to the Vertex channel

If your use case strictly requires a pure white background (e.g., e-commerce product shots), routing through Vertex fixes the root cause.

​
VertexGemini group reference

​
How to use

1

Request access from support

Contact APIYI support / ops to enable the VertexGemini group. It is currently invite-only.
2

Create a dedicated token

Log in to the APIYI console → Token management → Create token → select group VertexGemini → save the key.
3

Switch the call

Use the new token to call Nano Banana Pro / Nano Banana 2. The Base URL stays the same (https://api.apiyi.com/v1).
The VertexGemini group has limited concurrency (~100 RPM), so it is not recommended for high-throughput production workloads. Keep Aistudio as your primary channel and only switch to Vertex for pure-white-background testing.

​
Common Questions

The Aistudio rendering issue is triggered by compressing large uniform-color regions. Switching to light gray or beige gives the background some texture, which avoids the “compress a large uniform area” condition that triggers the artifacts.
Not necessarily. Vertex is more stable on pure-white / uniform-color backgrounds, but on other scenes (complex composition, portraits) the two are comparable — and Vertex is slower. Only switch when you have confirmed the issue is triggered by pure white.
It is invite-only — contact APIYI support or your account manager. For most users, Option 1 (switch to a light-tint background) is sufficient.
No. The Base URL stays https://api.apiyi.com/v1. Only the token’s group changes; APIYI’s backend routes Vertex-group traffic to the Vertex platform automatically.

Nano Banana Image Failure

Common issues and troubleshooting for Nano Banana image models.

Google Aistudio vs Vertex Channel

Differences between the Aistudio and Vertex Gemini image channels and how to choose.

Enterprise Vertex Fallback Group

How enterprise customers use the Vertex group as a fallback channel.

Image Async API

How to use the asynchronous image generation endpoint.