Interpreting is a very difficult job, and it is known that human interpreters are advised to interpret at maximum about 20 mins at one go. So, is there a possibility to develop a machine tool with the power of multiple interpreters simultaneously translating to more than one language? During the Open Doors Day at ÚFAL (November 21, 2019) we showed a prototype of live machine subtitling.
Of course there are problems, some of them rather expected, such as the lag in German due to a greater volume of words or mistakes due to non-standard accents of non-native English speakers or various paces of speech. But the goal of machine live subtitling at conferences is getting closer.
Here is a short video: