The data encrypted in the migration patterns of beluga whales indicate the dependence of biological species on a variety of external factors, including natural disasters, human actions and technologies. Training of technological agents can help preserve this data and present it as an aesthetic experience. The LSTM recurrent neural network is trained on data of the movement of beluga whales in coastal waters of Alaska from 1979 to 2010 and generates coordinates of alternative migration routes for beluga whales, which becoming the basis for animation. Data on the depth and temperature of the water at the current point are reflected in the sound of the work through parametric sonification of the synthesized sound and audio recordings of beluga whale communication using another neural network - a multilayer perceptron.