ReFrame

Inspiration

Photographers often take multiple shots of the same scene with slightly different lighting, angles, and exposures, but only one ends up getting used. We were inspired by how platforms like AutoHDR process images at scale, and wondered: what if we could automate not just enhancement, but also selection? ReFrame was built to eliminate the manual effort of choosing the best shot and editing it, turning a messy batch of photos into a single, polished result instantly.

What it does

ReFrame allows users to upload multiple similar images of the same scene. It then:

Analyzes each image for quality metrics like sharpness, brightness, contrast, and composition

Selects the best candidate automatically

Applies HDR-style enhancement to produce a final, high-quality image

The result is a streamlined pipeline that mimics professional real estate editing in seconds, with no manual intervention.

Challenges we ran into

One major challenge was reliably selecting the “best” image. Simple metrics like brightness or sharpness alone were not sufficient, so we had to combine multiple signals into a weighted scoring system.

Another challenge was making HDR enhancement look natural. Over-processing often led to unrealistic colors or halos, so we had to carefully tune parameters to balance clarity with realism.

Handling multiple large image uploads efficiently without slowing down the system also required optimizing our backend processing pipeline.

Built With

• cloudinary
• featherless
• tailwind