Skip to main content

​
Quick Answer

The Nano Banana series (including Nano Banana Pro and Nano Banana 2) uses Google Gemini models under the hood. The primary reason for image generation failures is triggering Google’s content safety mechanisms, which block non-compliant requests during the generation phase. API Yi acts as a transparent proxy, faithfully forwarding Google’s feedback.

​
Common Triggers

NSFW Content

Prompts or reference images containing pornographic, violent, gory, or hate speech content are directly blocked by Google’s safety mechanisms.

Watermark Removal

Requesting watermark removal from images violates content policies and will be rejected by Google.

Famous IP / Copyrighted Characters

Generation requests involving copyrighted characters from Disney, Marvel, Nintendo, etc. are rejected due to copyright protection.

Minor-Related Content

Any inappropriate content generation requests involving minors are subject to Google’s zero-tolerance policy and strictly blocked.

​
New Restrictions in Nano Banana 2

Since Nano Banana 2 launched on February 27, 2026, Google has further tightened its safety mechanisms. The following scenarios also trigger blocking:

Famous Individuals

Image generation or editing requests involving public figures (celebrities, politicians, etc.) will be rejected.

Financial/Order Info Modification

Attempts to modify financial information, order screenshots, price tags, etc. in images will be blocked.

Outfit/Face Swapping

Operations like changing outfits or swapping faces on people involve privacy and ethical concerns and will be rejected.

Implicit NSFW Content

Even without explicit sexual descriptions, suggestive or borderline inappropriate content will be detected and blocked.

​
Policy Update Timeline

DateEventImpact
January 23, 2026Google adjusts risk control policiesOverall safety review becomes stricter; some previously allowed prompts are now blocked
February 27, 2026Nano Banana 2 launchesNew blocking rules for famous individuals, financial info modification, outfit/face swapping, implicit NSFW content

​
Typical Symptoms of Generation Failure

When image generation fails, the API still returns HTTP status code 200, but the response contains no image data — instead, it returns a text explanation.
Why is the status code 200? API Yi acts as a transparent proxy, faithfully forwarding Google API’s original response. Google returns a 200 status code with a text rejection explanation during safety blocking, rather than an HTTP error code.

​
Common Error Messages

Google API’s rejection text typically contains messages like:
  • "I can't complete the modification of xxx"
  • "I can't generate images that are sexually explicit."
  • "I'm just a language model and can't help with that."
Note: Google’s safety filtering has some randomness. The same prompt may succeed sometimes and fail other times, depending on reference image content, prompt combination, and other factors.

​
Recommendations for Product Developers

If you’re building user-facing products on the Nano Banana API, we recommend implementing proper error handling logic to provide friendly failure messages to your users. We provide a detailed error handling development guide with complete decision flows, code examples, and user-friendly message templates: 📖 Error Handling Guide (Feishu doc): xinqikeji.feishu.cn/wiki/Rslqw724YiBwlokHmRLcMVKHnRf
Key Indicators to Check:
  1. candidatesTokenCount = 0: Google rejected at the content review stage
  2. finishReason is not STOP: Blocked by safety policy during generation
  3. Text present but no image: API returned a rejection explanation instead of image data

​
Contact Support

If your use case is legitimate and compliant but you’re still experiencing generation failures, feel free to contact us for investigation:

Technical Support