Local task records

Source copies, results, diagnosis, and QA evidence persist on the computer running the local service.

External AI processing

Before generation, the product discloses that images may be sent to the configured external service and waits for confirmation.

Review before delivery

Comparison tasks require a preferred result while QA risk and output evidence stay visible.

Source diagnosis

Decide where AI can help and where it must stay uncertain

Recoverable source signal

Face outlines, grayscale differences, hairlines, clothing edges, and scene structure determine whether enhancement remains evidence-based.

Small faces, severe damage, missing features, and valuable family portraits need more conservative comparison and manual review.

Generation begins only after disclosure that the configured external AI provider may receive the source and necessary processing images.

Restoration evidence

Key principles

Damage, blur, exposure, photo type, recoverable signal, and identity risk guide the initial restoration route.

A, A+, B, and C expose different restoration strengths so the clearest output is not accepted automatically.

QA checks faces, people count, composition, crop, borders, background reconstruction, over-sharpening, and vintage texture before delivery.

FAQ

Does AI recover the exact missing detail?

Not necessarily. It can infer plausible content, so weak or missing evidence must not be described as certain historical recovery.

Are photos processed fully offline?

No. After explicit confirmation, images can be sent to the configured external provider for diagnosis, restoration, and QA.

Why keep several AI versions?

Comparison makes identity drift and over-enhancement easier to detect than reviewing one polished output in isolation.

Ready to begin

Upload the source, choose a restoration route, and inspect likeness, composition, and quality evidence before delivery.