AI and digital editing rules for US passport and visa photos

Quick answer

Do not use AI generation, face replacement, filters, or retouching for an official US passport or visa photo. Use a real uploaded photo and limit preparation to crop, sizing, and measurable quality checks.

The product rule

For US passport and visa photos, the product rule should be simple: do not generate the official image with AI.

That does not mean software cannot help at all. It means the software must stay in a preparation role:

• crop the uploaded image
• resize the output
• check that only one face is present
• estimate whether the face is large enough
• warn when the source image is likely unsuitable

It should not invent missing details, smooth skin, reshape the face, replace the applicant, or create a studio portrait from references.

Why AI generation is different from crop/export

A crop keeps the submitted image anchored to the real upload. AI generation creates or modifies image content. For official US photos, that distinction matters.

Examples of risky changes include:

• face retouching
• background replacement
• changing expression
• changing hair or skin appearance
• combining multiple reference photos into one result
• generating a passport-style headshot from a selfie

Even when the result looks realistic, it can violate the official restriction.

What to do instead

The safer workflow is stricter but clearer:

1. ask the user for a real photo
2. verify that exactly one face is detectable
3. crop to the US square output
4. export a digital file
5. explain that this is a measured preparation, not an approval guarantee

If the uploaded photo is not good enough, the service should ask for a better source photo rather than creating a synthetic correction.

How this affects YapaPhoto internationalization

France and US flows should not share the same generation promise. A France flow may have its own compliance strategy, but the US flow must be isolated:

• separate route
• separate guide content
• separate workflow standard code
• no AI-generation call for US official output
• checkout and print products only after a dedicated US product decision

That separation protects users from a misleading output and keeps the implementation easier to audit.

Recommended method

1. 1
   Choose a compliant source photo

   Start with a real image that already has suitable lighting, background, and expression.

2. 2
   Limit changes to preparation

   Crop and resize without changing the face, background content, or apparent identity.

3. 3
   Keep acceptance wording precise

   Present automated output as preparation and precheck, because official acceptance is decided by the reviewing agency.