123RF has introduced a generative AI-driven video comprehension capability built on Amazon Web Services (AWS), marking a major shift in how users search for and license creative assets. The Malaysia-based stock media platform, which hosts more than 230 million assets for 12.4 million users worldwide, unveiled the upgrade during re:Invent on 3 December. The company aims to move beyond traditional keyword-dependent search by using AI to understand the actual visual content within videos and images.
The new system uses Amazon Nova to interpret visual elements directly, helping the platform improve the accuracy of descriptors and search relevance. Early testing on a dataset of five million videos delivered twice the accuracy in video descriptors and enabled more precise content matching. This approach is particularly beneficial for ecommerce businesses that depend on rapid and accurate asset selection, allowing them to pair visuals with product listings more efficiently and increase daily output.
Conventional search often returns irrelevant results when metadata is incomplete or incorrect. The platform’s latest upgrade reduces this issue by analysing visual content instead of relying on tags. When a user searches for a green bag, for example, the system identifies genuine images or videos featuring green bags, rather than unrelated assets that include the word in their metadata or contain a green background. The AI can also detect trademarked patterns, logos, and branded items, enabling automated flagging and correct licensing categorisation.
Bernadine Michael, Chief Marketing Officer at 123RF, said the adoption of AWS technologies has reshaped the way the company operates. “AWS has fundamentally changed how we operate at 123RF, transforming us into an AI-powered creative enabler. Our marketing teams now launch campaigns 35% faster. This efficiency means our teams can focus less on repetitive work and more on serving the creator economy,” she said. She added that the real impact lies in democratising creativity across its global subscriber base, helping designers and marketers find the right asset instantly, regardless of language or cultural context.
Streamlining compliance and multilingual content
123RF processes more than three million image uploads each month, each requiring compliance checks, copyright verification, and quality assessment. The workload previously required a team of 30 to 40 human reviewers handling around 3,000 images per day. The platform also needed to deliver more comprehensive multilingual support, as most metadata was submitted in English even though the platform supports 15 languages. The challenge extended beyond translation to understanding idioms, cultural context, and nuance.
The adoption of AI-driven visual understanding and automated checks has reduced the overall content review time by 92%. Customers can now discover relevant assets in 90% less time, with improved accuracy and fewer irrelevant or duplicate results.
Phoebe Liew, Chief Technology Officer at 123RF, said the new system allows the platform to interpret images in ways that closely resemble human perception. “Our AWS-powered AI technology now ‘sees’ images the way humans do,” she said. “When someone uploads an image, our system instantly captures what makes it visually unique – like identifying a specific architectural style or recognising the composition of a sunset scene. This visual understanding means we can match similar images automatically, regardless of what language the description is in.” She said this approach reduces dependence on contributor-generated keywords and helps prevent copyright issues by examining visual elements directly.
Building a scalable global infrastructure
To support its expansion, 123RF used several AWS generative AI services. The company leveraged Amazon Bedrock with Anthropic’s Claude 3 Haiku model to build a translation system designed to handle cultural nuance. This allows users to avoid misinterpretations, such as differences in the meaning of terms like “freedom” in English and “jiyū” in Japanese. The system can also identify quality issues, copyright icons, and logos without manual intervention.
Amazon Nova Pro and Nova Lite understanding models help the platform assess copyright risks before content goes live, offering another safeguard for users who rely on the accuracy and legality of creative assets.
Hussein Mohd. Ali, Country Manager for AWS in Malaysia, said the collaboration demonstrates how AI is reshaping creative workflows. “123RF demonstrates how AI transforms creative workflows – tasks that once took designers days now take minutes, enabling Malaysian companies to compete on a global stage,” he said. He added that findings from AWS’s Unlocking Ambitions survey show that nearly half of startups are already implementing AI solutions, with almost a third developing entirely new AI-driven products.
The company believes its latest enhancements will set a new benchmark for the content licensing industry and help creators, marketers, and businesses work more efficiently through AI-driven content discovery.
