The Translating Telephone allows several users speaking different languages to communicate. Each user receives translated text and optional speech synthesis in his or her own language. The app was built using WebRTC for video conferencing, SYSTRAN or Bing for machine translation, and the Web Speech API for speech recognition and speech synthesis. Later, a similar idea was incorporated into Skype.
The big and obvious problem is that errors in speech recognition are amplified when they are followed by translation, which has its own error rate. A good result requires much effort on the part of the speaker.