New technologies have made contemporary communication easier than ever. According to a recent Apptopia survey, today’s top four apps in terms of screen time are all social apps. In 2015, half of all Instagram comments included an emoji. By 2017 over 60 million emojis were being sent daily on Facebook, and over 5 billion on Messenger. Almost all of those were stock platform emojis. As the use of and dependence on social apps continues to grow, so have attempts to personalize the experience, and this has engendered a thriving emoji industry.
Stock emojis and emojis designed by artists are fun, but lack the personalisation many are seeking in the expanding emoji marketplace. More and more social apps are now leveraging AI technologies such as machine learning, deep neural networks and facial recognition to enable users to create highly-individualized emojis from selfies and photos.
Stanford CS Project: use deep learning to create image + word stickers
- A couple of Stanford University students who are emoji and AI fans developed a sticker-generating system for their CS224n (Natural Language Processing with Deep Learning) final project. More than 400,000 “image + word” style stickers were used to train a deep neural network. They also proposed two indicators to assess quality and human differentiability, with results showing the AI-generated stickers were almost as humorous as human-created stickers. The students say one of the biggest challenges in developing their deep learning model was understanding the humour of different groups and cultures. They also committed to further research designed to eliminate potentially sensitive or offensive output.
Mirror AI – Create cute Stickers featuring your face
- Mirror AI is a startup founded in 2017 in San Francisco which developed an app to allow users to transform selfies into personal cartoon emojis by capturing facial features with a built-in image processing tool. Along with various clothing and background options, the engine also enables personalising makeup, hairstyle, and even facial expressions. It is compatible with all major social apps and supports creating personalised sticker packs and texts, as well as co-generating emojis with friends.
Microsoft swiftkey ‘puppets’ for 3D animated emojis
- Microsoft SwiftKey “Puppets” brings advanced AI technology into play. Puppets are AR-based animated emojis that incorporate users’ facial expressions in the form of virtual characters that are as cute and clean as Disney animations. The system was trained to track and record muscular movements using a dataset of videos and images of facial expressions. Currently, there are five different animated characters available for transformation, and users can generate clips of up to 30 seconds.
Emojis are evolving. Startups and established social tech companies alike are hiring designers and emoji artists, and popular emoji IPs are even spawning peripheral products. All the signs indicate the strong potential of this new industry. With the introduction of AI technology, emojis can now be highly personalized by users with only a few clicks, bringing an unprecedented level of individuality to online chats and interactions.
There are however a number of challenges. Emojis are largely used for entertainment, and so ideally the neural network that generates them needs to understand something about human humour — and that has proven a daunting research challenge. Additionally, the security of the facial-feature analyzing systems in these apps remains unproven, which could result in privacy breaches involving the biological date they collect and store on smartphones and remote servers.
AI has brought fun, convenience, and personalisation to today’s personalised and dynamic emojis. In the future, we may see emojis built with advanced visual technologies reach wider areas, such as digital marketing, film and TV shows. We’ll have to see where our emoji culture takes us next!
Author: Ziyang Lin | Editor: Michael Sarazen