I'm interested in accessing the users audio for authentication analysis. I am open to more involved and technical solutions that require custom servers. It also does not have to be on the skill store. It is just for a proof-of-concept demonstration.
I know this has been "answered" before. Everyone seems to think it is only possible to get text/json from the user's utterance and not the raw voice mp3 that Amazon parses for us. But, with a custom server solution, it seems like it is possible.
If you check the video at the bottom of the page here. He says Alexa is hooked up to their own "custom servers" instead of Amazon's. How is this possible? Any pointers?