• by Laurie Sullivan  @lauriesullivan, September 22, 2015

The Finnish startup Valossa has developed a video voice search technology based on artificial intelligence (AI) that developers and original equipment manufacturers can integrate into their platform to compete with companies such as Apple. 

The AI technology gives voice to search engines that power queries on TVs, mobile devices, and PCs. It recognizes the content of video files through a combination of text and pattern recognition, and can reach into the video to identify the context through metadata, making it searchable. It also can discover and recommend semantically-related content from large video collections.

Mika Rautianinen, Valossa co-founder, says future query features will enable users to drill down deeper into content. "By having scene-level understanding of the audiovisual entities in content we can better contextualize in-roll and pre-roll ads," he says. 

Technologists at Valossa plan to train the machine learning system to analyze more complicated characteristics such as moods and use of special effects. For instance, "find me the videos having highest tension and lots of special effects" vs. "I would like to see something soothing in an Italian setting." The results for these queries would consider analysis of full video content characteristics, including emotional tones and what elements appear in the audiovisual stream.

Engineers at Valossa, spun off from the University of Oulu, made the technology compatible with Mac OS, Microsoft Windows and Linux. While the technology integrates with Amazon, Google, and Microsoft, the server-side technology enables nearly identical functions to the Apple TV OS for any system provider. 

Still, Valossa will need to compete with Apple TV, which now integrates what many believe is a very good revision of Apple's voice assistant Siri. "What we saw with Apple demos can be done easily with superficial metadata," Rautianinen says. "They announced simple filtering by genre, cast, director and age rating."

Valossa already goes beyond this capability, allowing users to describe objects, events, characters, quotes or locations they seek in a movie. He says voice search queries like "Find me romantic movies in Hawaii" give you the best fitting matches based on actual content modeling, not just simple matching of search words with editorial metadata. 

The goal will be to bring Valossa's technology to companies competing with Apple TV. Competitors can integrate the artificial intelligence and search capabilities. "Apple doesn't license its TV platform to others, so competitors need to bring their features up to a new standard," Rautianinen says. "Companies like Rovi have had their voice search years before Apple. Apple is not alone in the domestic entertainment space and the race for the supremacy in voice controlled entertainment has just started."