Google Announces 2.2MB Version Of Voice Recognition Algorithm VoiceFilter

Google Announces 2.2MB Version Of Voice Recognition Algorithm VoiceFilter

Google announced the version of its voice recognition artificial intelligence called “VoiceFilter“, which was announced in 2018. The file, which is only 2.2 MB in size and can work with tons of applications, prevents the difficulties of sound perception. However, this artificial intelligence only supports the English language for now. US-based technology giant Google has developed various services throughout the years. Speech recognition features are among these services. For example, all features such as Assistant, Translation and Voice Recorder found in Translation and Pixel devices provide speech perception in some way. However, getting it right is a difficult issue. Aware of this situation, Google has now announced a new artificial intelligence model that makes things easier.

The biggest problem with applications that can recognize speech was to extract users’ voices from other speakers and noise in the environment. The company took action to overcome this problem in 2018. Google created an artificial intelligence called VoiceFilter and it was working too slowly. Google, which remodeled this artificial intelligence, is now introduced to consumers with “VoiceFilter-lite“, which occupies only 2.2 MB of space on devices.

Google Announces 2.2MB Version of Voice Recognition Algorithm VoiceFilter
Google VoiceFilter

Artificial intelligence models for voice recognition lead to higher battery consumption and CPU usage. Google‘s new artificial intelligence model prevents this. The artificial intelligence model, which can work in harmony with tons of mobile applications, will provide various advantages to both Google and other developers.

According to some statements made by Google, VoiceFilter-lite uses the user’s already analyzed voice. However, if there is a conflict during detection, artificial intelligence works to prevent this conflict. In addition, this artificial intelligence model does not need to be integrated into the main model. Naturally, a previously undefined voice can bypass VoiceFilter-lite and switch to a different voice recognition model. Artificial intelligence, which can only be used in English for now, is expected to support different languages ​​in the future.