AI Transparency
How our AI works
Last updated: 14 April 2026
What we use AI for
Studio Hair uses generative artificial intelligence to (1) analyze your face shape, hair texture, and complexion from a selfie you upload, and (2) generate synthetic hairstyle preview images of you with various haircuts.
This is the entire purpose of the AI in our service. We do not use AI to make decisions about you, score you, profile you, or predict anything about your behavior or characteristics.
The models
- Face validation:
gemini-2.5-flash(Google Vertex AI, EU region) — confirms the uploaded image is a usable face photo. - Style selection & prompt writing:
gemini-3.1-pro-preview(Google Vertex AI) — analyzes your face and selects suitable hairstyles from our curated catalog using a 5-gate professional-stylist decision framework. - Image generation:
gemini-3.1-flash-image-preview(Google Vertex AI) — generates each hairstyle preview image.
These foundation models are trained and maintained by Google per their AI Principles and model documentation at ai.google.dev/docs/models. Studio Hair does not train, fine-tune, or modify these models in any way — we provide the input (your selfie plus a prompt) and return the output to you.
What the outputs are
Generated images are synthetic visualizations — not photographs of you. The AI imagines what you might look like with a given hairstyle. They are intended for personal entertainment and pre-salon consultation purposes only.
AI image generation is imperfect. Outputs may contain artifacts, may not be perfectly realistic, may not match what a real salon could achieve, and may not be flattering. They should never be presented as actual photographs of you.
EU AI Act compliance
Studio Hair's AI system is classified as limited-risk under the EU AI Act (Regulation (EU) 2024/1689). It generates synthetic content depicting natural persons; we therefore comply with Article 50 transparency requirements:
- Article 50(1): we tell you clearly that you are interacting with an AI system. You see this on the generation page (persistent banner) and at signup.
- Article 50(2):generated outputs are marked as AI-generated in their metadata and visibly labeled (“AI-generated by Studio Hair”) in the user interface.
- Cryptographic provenance (C2PA): implementation in progress; will ship before the EU AI Act full enforcement deadline of 2 August 2026.
What the AI does NOT do
- Identify you to anyone
- Profile you, score you, or rank you
- Make automated decisions about you (GDPR Article 22 does not apply)
- Train on your selfie or your generated images
- Share your data with anyone outside our processors (Google, Stripe)
- Predict your behavior, emotions, or attributes beyond hair-relevant analysis
- Process selfies of anyone other than yourself (you must own and consent to use the image)
Where AI processing happens
All AI inference runs on Google Vertex AI in EU-region datacenters (primarily Belgium). Selfies and prompts are processed in-region; no AI inference happens outside the EU.
Data retention for AI processing
- Selfies: deleted within 60 seconds of generation completing
- Generated previews: approximately 180 days (6 months) from creation, then auto-deleted by Google Cloud Storage lifecycle (sweeper runs asynchronously, ±24h), so you can re-download your Report during that window
- AI prompts (the text we send to the model): not retained beyond the request
- Face analysis output: stored on the job document for the same 180-day window so you can see why each style was recommended
Bias and limitations
Image-generation AI models can reflect biases present in their training data. Studio Hair's AI may produce better results for some face shapes, hair textures, and skin tones than others. We use a curated 142-style catalog explicitly covering all six face shapes and four hair-texture types (Andre Walker scale) to mitigate this.
If you notice the AI consistently underperforms on your features, please email contact@studiohair.app. Your feedback helps us improve catalog coverage.
Human oversight
Our pipeline includes automated quality checks (post-generation validation) but we do not manually review every generated image before delivery. You remain the final arbiter of whether to use a given output.
Your rights regarding AI processing
You can withdraw your consent to biometric AI processing at any time from your account settings. After withdrawal, future generations require re-consent and any cached face analysis is deleted.
See our GDPR rights page for the full set of rights, including objection to processing and erasure.
C2PA cryptographic watermarking will be added before the EU AI Act enforcement deadline (2 August 2026).