One-Day AI Photo Restoration Test

Overview

This short project explored how quickly old photographs could be restored and colorized using AI tools. In just one day, multiple tests showed that consumer-level AI platforms already deliver surprisingly strong results for enhancing vintage images.

Concept

The idea was to test the limits of AI-powered photo restoration on a fast timeline, seeing how well images could be cleaned up, sharpened, and colorized.

Tools

• Runway ML – For restoration and colorization

• Nano Banana AI (Google) – Strong performance in photo enhancement and colorization

• Photoshop – For close-up evaluation and potential refinement of AI outputs

Process

1. Collected old photographs for testing

2. Ran them through Runway ML’s restoration and colorization pipeline powered by Nano Banana AI

3. Brought results into Photoshop to inspect quality up close and assess how much further they could be refined with manual work

4. Compared outcomes across several tests to evaluate speed and quality

Reflection

At a consumer level, the results were excellent, fast, intuitive, and visually impressive. With additional refinements in Photoshop, these outputs could absolutely be turned into beautiful restorations.

That said, close inspection reveals a crucial point: AI-based upscalers and enhancers sometimes generate unwanted or inaccurate details. In one test, for example, the restored image showed a ring on the subject’s left hand that wasn’t present in the original photograph. Similar artifacts appeared across different images.

This highlights the need to pay close attention to detail, especially in professional contexts like historical photo restoration. While likeness and facial features are often preserved remarkably well, when viewed up close, the technology can “fall apart.” As powerful as it is, AI is still not perfect, and human agency and verification remain critical to ensure accuracy and authenticity.

From a professional perspective, AI might not fully replace skilled restorers, but it could dramatically speed up workflows. And when thinking about AI replacing jobs, this is one of those cases where it may genuinely shift how restoration work is approached.

Output

A series of restored and colorized photographs completed within a single day.