Why Gemini Won't Generate Images of People (Nano Banana)
Why Gemini and Nano Banana refuse to generate or edit photos of people, including your own: age rules, filters, false positives and fixes.
Gemini (Nano Banana) usually refuses a person image for one of four reasons: the photo looks like it includes a child, the request reads as sexual or revealing, it involves a real person in a way that could deceive or harass, or your account does not meet the age requirement for editing (18+). Several of these checks are cautious by design, so ordinary selfies, family photos and swimwear shots get blocked more often than the policy strictly requires.
Disclosure: VdoBloom is our platform.
Nano Banana is the name Google uses for the image models inside the Gemini app (Nano Banana 2 for everyday use and Nano Banana Pro for paid plans, as of September 2026). They are very good at editing a photo while keeping a face recognizable, which is exactly why Google filters person edits carefully. Below is what Google's published rules say, why normal requests get caught, what you can change, and which other models make sense when Gemini keeps saying no.
What Google's rules say about images of people
Gemini image generation is covered by three official documents: the Gemini Apps help page for creating and editing images, the Gemini app policy guidelines, and Google's Generative AI Prohibited Use Policy. Together they say:
- Age limits. You must be 13 or older (or your country's minimum) to generate images, and image editing is not available to users under 18. If your Google account has no confirmed age, or it is a school or supervised account, editing may simply not work.
- Child safety. Gemini should never generate content that exploits or sexualizes children.
- Sexually explicit material. No explicit sexual content and no depictions of sexual violence. The Prohibited Use Policy also bans facilitating non-consensual intimate imagery.
- Impersonation and deception. Google prohibits impersonating a real person, living or dead, without disclosure in order to deceive, and using personal data or biometrics without legally required consent.
- Violence, harassment and hate. No gratuitous gore, and nothing that incites violence against or dehumanizes a person or group.
- Your uploads. Google reminds users to make sure they have rights to any image they upload and not to violate others' privacy.
Google also says context matters and that it weighs educational, documentary, artistic and scientific uses. In practice the automatic filters cannot always see that context, which is where most false refusals come from. Every image Gemini makes also carries an invisible SynthID watermark, so edited photos can be identified as AI-made.
Why Gemini won't generate or edit my own photo
The frustrating case is when you upload a picture of yourself and ask for something harmless. These are the most common causes.
The photo looks like it includes a minor
Filters for minors are deliberately cautious across the industry. A youthful face, a school-style outfit, or a child anywhere in the frame (including the background or a photo on the wall) can stop the whole edit. Cropping to just yourself fixes many of these.
Your account cannot edit
Because editing is 18+, a Workspace or school account, a supervised account, or an account without a verified age may get refusals on every upload while text-only generation still works. Try a personal account where your age is set.
Clothing, body and pose requests
"Put me in a bikini", "make the dress shorter" or "change my body shape" edit how much skin a real, identifiable person shows. Even when the result would be ordinary beachwear, Gemini is conservative about it. We covered the swimwear case in detail in why Nano Banana blocks fashion and swimwear images.
Photos of other people
Editing a stranger, an ex, a coworker or a celebrity into a new situation touches the impersonation and privacy rules. Gemini cannot know whether that person agreed, so it errs toward refusing, especially for realistic results.
Wording that implies harm
Words tied to injury, weapons, fighting, or making someone look "drunk", "arrested" or "ugly" can read as harassment when a real face is involved, even if it is a joke about yourself.
Quota and product limits
Not every failure is a policy decision. Gemini has daily image quotas, and Google's help page notes that once you hit the Nano Banana 2 quota you cannot redo images with Nano Banana Pro either. If simple requests also fail, you may have hit a limit.
What you can legitimately change
If your request is allowed and got misread, these adjustments help. They are about giving the model accurate context, not about sneaking prohibited content past it; attempts to disguise prohibited requests break Google's terms.
- Crop to adults who agreed. Remove children, bystanders and anyone who has not consented from the upload.
- State what the image is for. "Professional headshot for my LinkedIn profile, same face, navy blazer, soft studio light" is clearer than "make me look hot."
- Describe clothing as a product. "A one-piece navy swimsuit, catalog style, standing on a beach" reads differently from adjectives about the body.
- Keep the face unchanged and say so. Asking to "keep my face exactly as it is" signals an edit of yourself, not a new fake person.
- Use a personal, age-verified account for photo edits.
- Start a new chat if earlier messages in the thread were refused; the conversation history can influence later answers.
Other models for editing photos of people
Nano Banana is available on VdoBloom too, but it is badged STRICT there because Google's filters still apply. Moving to VdoBloom does not change what Nano Banana will accept. If Gemini is refusing an ordinary adult request, the useful switch is to a different model family. In Image Edit you upload your photo, describe the change, and choose a model:
| Model | Badge | Credits per image |
|---|---|---|
| Nano Banana Edit | STRICT | 5 |
| Nano Banana 2 | STRICT | 15 |
| Nano Banana Pro | STRICT | 18 (24 at 4K) |
| Auto — Smart Edit (Seedream 4.5 with WAN 2.7 fallback) | HOT | 7 |
| Seedream 4.5 Edit | HOT | 7 |
| Qwen Image 2.1 Edit | HOT | 3 at 1K, 5 at 2K |
| WAN 2.7 Image | HOT | 4 |
HOT means fewer restrictions on adult, clothed, non-explicit content such as swimwear, fitness and fashion edits. It does not mean anything goes. VdoBloom screens every prompt before charging and blocks nudity and undressing requests, and it always blocks sexualized content involving minors. If a model's provider blocks a result after the job starts, your credits are refunded.
Two worked examples: five outfit variations of your own photo on Seedream 4.5 Edit cost 35 credits; the same five on Qwen Image 2.1 Edit at 1K cost 15. New accounts get 10 free credits to try a few edits.
For specific jobs there are dedicated tabs. AI Headshots turns one photo into styled professional portraits at 5 credits per image, and AI Fashion Try-On puts an outfit from one image onto the person in another, also 5 credits per image. Both use Nano Banana Edit, so they follow Google's stricter rules; they work well for professional and everyday looks, while the HOT models above are the fallback for swimwear or more stylized edits.
Consent comes first
Editing a recognizable face is where AI images can hurt real people, so the rules here are firm:
- Upload only your own photos, or photos of adults who have agreed to be edited. VdoBloom asks you to confirm that everyone pictured is 18+ and that you have consent before any upload tab will run.
- Never upload images of minors for any suggestive edit. VdoBloom blocks it.
- Never create intimate, sexual or undressed images of a real person. That is non-consensual intimate imagery, and it is a crime in many places.
- Do not place public figures in realistic scenes meant to make people believe something false.
Frequently asked questions
Why won't Gemini generate images of people?
Gemini does generate people, but it refuses when a request looks sexual, violent, deceptive, or involves someone who may be a minor. For uploads, editing also requires an account holder aged 18 or older. Clear, purpose-first prompts and cropped photos fix many refusals.
Why does Nano Banana refuse to generate the image?
Nano Banana applies Google's Prohibited Use Policy and Gemini's policy guidelines. Common triggers are revealing clothing on a real person, a child in the photo, a celebrity or other real person in a realistic scene, and wording that implies harm. Quota limits can also look like refusals.
Is Nano Banana censored more than other models?
It is among the stricter image models for person edits, which is why VdoBloom labels it STRICT. Models such as Seedream 4.5 and Qwen Image 2.1 are badged HOT because they accept more adult, clothed content, though they still block nudity and anything involving minors.
Can I use Nano Banana without Gemini's filters?
No. Using Nano Banana through another platform, including VdoBloom, still runs Google's filters. If your request is legitimate but Gemini rejects it, use a different model whose policy covers it instead of trying to work around Google's.
Why did Gemini edit my photo yesterday but not today?
Results are generated fresh each time and checked afterwards, so small differences can change the outcome. Google also updates its filters, and you may have reached your daily image quota.
Ready to try it?
Create your first AI video in minutes — no credit card required.
Start Creating Free →