A British photo printing service faced a familiar problem. Customers often struggled to name their digital albums, leaving them with generic labels like Vacation 2023 or Untitled Folder. Popsa turned to Amazon’s artificial intelligence tools to solve this. The company integrated Amazon Bedrock and models from the Amazon Nova family into its existing system. This upgrade automated the generation of personalized titles and subtitles for photo collections.
Popsa combined three data sources to create relevant suggestions. It used image metadata, computer vision analysis of photos, and retrieval-augmented generative AI. The system now produces creative titles that match each brand’s voice. It supports twelve languages, including English, German, French, and Spanish. The update covers all customer photo books and digital albums.
The technical shift involved three specific models. Popsa adopted Anthropic’s Claude 3 Haiku for fast, lightweight processing. It also used Amazon Nova Lite for cost-efficient suggestions and Amazon Nova Pro for higher-quality outputs when needed. All three models connect through Amazon Bedrock’s unified API. This setup allowed Popsa to balance speed, cost, and quality without switching providers.
Results showed clear improvements. Response times dropped significantly as the AI handled more requests per second. Operational costs fell by reducing the need for manual title reviews. Customer feedback indicated higher satisfaction with automatically generated names. Popsa now refreshes titles in real time as users add or remove photos.
The company plans further refinements. It will expand language coverage and fine-tune suggestions for specific photo types. Popsa aims to make album naming effortless while keeping each title unique and brand-consistent.
Source: aws.amazon.com