Auto-generating detailed information about your video assets
Artificial Intelligence was one of the key themes of IBC 2017, and Valossa AI was one of the products that stood out, due to its range of AI services and its clean, consumer-style interface.
It takes many of the main AI functions – face recognition, dialogue analysis, mood analysis – and packages them into a single product/API to enable content owners to extract invaluable information to help monetise archives.
Finnish company Valossa is less than two years old and grew out of one of Europe’s leading computer science and AI labs at the University of Oulu. It has just completed a US$2m round of funding, to help expand its product team and its global sales and marketing, so it clearly aims to be a sizable player in this space.
Its product Valossa AI analyses videos, automatically generating descriptive tags, categories and overviews. It can be used to help locate and search your archives for anything (dialogue, emotions, colours, objects, etc) inside a video.
It recognises faces and identities and can detect individuals and groups of people. It recognises what it sees in video clips – objects, settings, events and activities – and identifies sounds, music styles and human voices.
In addition, Valossa AI extracts keywords and topics from speech or transcripts and can categorise the themes of each clip. It also recognises and tags visual nudity and spoken profanity.
The demo I was shown at IBC 2017, of the latest version of Valossa AI, also recognised facial expressions, giving a percentage figure for each emotion, such as anger, fear, sadness, happiness, heart rate, surprise and so on.
To find out more about Valossa, and run a video clip through Valossa AI, click here. The first 15 minutes of analysis are free, then the cost is €99 for every five hours of video you want to analyse.
I ran a five-minute music documentary from artbeat.tv through Valossa AI – a few of the screen shots from the auto-generated analysis results are below.