Google claims to have developed a new translation system: Translatotron, Through the mechanism of artificial intelligence (AI) You can translate voice directly from one language to another without having to convert the text..
The California Department of Artificial Intelligence, a giant in Mountain View, California, has announced the new system. New alternative model from sequence to sequence Translation no longer requires intermediate text.
Operation of Translatotron It is based on a single process.Instead of dividing it into multiple stages, as in today's translation systems, which rely on a speech synthesis system,
With this system, we are confident that our researchers can achieve faster translation speeds. Avoid errors in speech recognition and translation., "I keep the original voice after translation."
This is how Google manages reach, as detailed in our official blog. You can better manage untranslated words., With the proper name.
How Translatotron Works
Our system uses the captured speech signal as the source, uses the spectrogram, and generates another voice line in the language you selected for translation. Also Neuro-speech coder Converts the resulting spectrogram to an audio wave with a time reference.
You can also add an additional mechanism to later learn the human voice characteristics and encode to preserve the voice tone for use in synthesizing voice translations.
In this process, Google's AI Multitasking goals to predict movement At the same time, it generates a spectrogram of the translation.