Qosmo updates “Imaginary Soundscape”, a web service where AI selects the soundscape for images and scenes
Qosmo Offers License for Updated Multimodal AI Technology for Commercial Purposes
MEGURO, TOKYO, JAPAN, April 28, 2022 /EINPresswire.com/ — Tokyo, Japan, April 27, 2021 – QosmoInc., which advances mankind’s creativity with art and technology, today releases the new updated version of “Imaginary soundscape”, an online service that finds an audio clip that matches uploaded images using multi-modal deep learning technology. This service is now offered free of charge in English and Japanese. Qosmo is also announcing the commercial license of the service’s core technology, Img2Sound Engine.
■ What is Imaginary Soundscape?
We all have the ability and experience to imagine the sound you would feel when seeing a photo. A beach would remind you of the sound of the waves, the image of a lively crossing would remind you of the sound of cars. This project is an attempt to externalize this human capacity. Using the image uploaded by the user, the AI model will find the best matching sound from the large sound library which consists of more than 60,000 sound clips. In the Google Street View mode, the user can walk around the location of their choice and experience the soundscape “imagined” by the AI. Since its initial launch in 2017, this project has been very popular and has been used by more than 500,000 users to date.
■ Features of the latest version
There are currently three major updates to the service: improved model accuracy, expanded sound library, and improved user interface. Internally, the AI model based on object recognition technology was replaced by a multimodal AI model built with the contrastive learning technique. Along with the expanded sound library, these updates allow for more contextualized matching with an awareness of subtle differences in nuance.
■ Commercial license of “Img2Sound Engine”
Right now we are launching the commercial license of our core Imaginary Soundscape technology, Img2Sound Engine. This technology is an application of a Deep Learning technique called Contrastive Learning. The input image and candidate audio files are converted to vector representations, allowing similarities between different content types to be compared. This technique is also applicable in the other direction (sound to image) as well as other modalities such as text, music and video. Qosmo is experienced in multimodal AI technologies, as we have helped our customers integrate them into their products, services and projects on various occasions.
Imaginary soundscape project reference page(currently based on previous version, update planned soon)
■ Past exhibitions and awards
– Submitted and selected for NIPS2017 Machine Learning for Creativity and Design
– Exhibited “Imaginary Soundwalk” at Media Ambition Tokyo 2018
– Favorite Website Award (FWA) Site of the Day
– Posted on “Experiments with Google – AI Experiments”
■ About Qosmo, Inc.
Founded in 2009, Qosmo’s vision is to advance humanity’s creativity with art and technology. A group of artists, researchers, designers and programmers work together for the mission to open the frontier of creativity through the production of works of art and to help build a society where more people can work. , learn and live more creatively by developing AI tools for artists, businesses and individuals.