We built a new product that uses Google Vertex and Amazon Nano to generate large-scale multimedia assets, which include text photos and videos with human realistic voices. Technically, we used Java as a backend pipeline system and also used NodeJS which uses AI capabilities and it is deployed on GCP. Importantly these videos and the voices are completely generated based on various metadata of e-commerce products which are fetched in real time from Amazon products database. We improved the performance of the oral system and also cache them in order to persist the responses of AI and to use it subsequently.
