During today’s big Amazon event, where numerous new Alexa devices were revealed, an interesting new capability to existing devices was announced. Amazon announced that the 4th-generation Amazon Echo and the 3rd-generation Echo Show 10 will soon be able to process all voice requests locally on the device without needing to send any recordings to the cloud. This is a pretty feature and one that has not been achieved by any smart speaker.
This feature announcement was done in a matter of seconds with very little detail. All that is known is that the feature will be an option for the two aforementioned Echo models due to the custom Amazon AZ1 processor in those devices. Note that Amazon announced the AZ2 processor today in the Echo Show 15, so it’s likely that Echo device will have the capability as well.
The option will be coming “soon” and will likely only be available in the US, at least initially. Processing voice requests locally would solve the concern many people have with how voice assistants currently need to send voice recordings to the cloud to determine what is being asked. Even with local voice processing, the intent of what is requested would still need to be sent to the cloud at times to retrieve information, but at least the concern of private conversations being recorded due to false triggers would be eliminated.
In addition to the privacy benefits, local voice processing also has the potential to speed up how quickly smart assistants respond. The 2nd-gen Fire TV Cube, for example, supports local voice processing for a small subset of common requests, like “Alexa, play” and “Alexa, go home.” These requests, which are processed locally on the Fire TV Cube, are executed up to 4 times faster than on the 1st-gen Fire TV Cube, which has to send all requests to the cloud. It really is noticeably better to see a command get executed the second you stop talking, instead of waiting several seconds for the blue light to finish flickering while it waits for a response from the cloud.