What if we told you that there is an AI that can recreate any sound that exists on planet Earth? So there is. We are talking about the VALL-E tool created by Microsoft. capable of imitating all available sounds based on only three seconds of some sampling.
To achieve this incredible result, the AI was “trained” for 60,000 hours with English data and speech, using three-second samples of certain sounds to create their sounds. However, what draws more attention to the project is that VALL-E manages to replicate the tone and emotions of its targets (including unpublished speech), as well as copying sounds. Something that has never been done before.
In some tests, it is possible to notice how the AI can perfectly recreate certain sounds. However, in some other cases, the voices still have more robotic properties. This is normal considering the car was only recently released and still needs tweaking throughout its existence.
In addition, it is necessary to take into account that the source only uses three seconds of sampling to work. If there were more seconds, the feature would be more likely to produce even better results. However, we will have to wait for new tests and developments to know which paths artificial intelligence will follow.
While you’re waiting for updates, if you’re wondering how VALL-E works, Cornell University has created a document containing various tests performed with artificial intelligence. You can listen to them here: valle-demo.github.io.
Microsoft’s product is not available to the general public for now, as it has passed some tests. This is still a positive thing. technology can be used for fraud and malicious schemes in a way., creating believable sounds. In that sense, the Windows company will need to find a way to make VALL-E a safe tool.
Source: Tec Mundo
