Reconstruction

Below is a comparison of reconstruction quality over the past several months. Here the AI takes an utterance (input), compresses it and then reconstructs (output) it from the hyper-compressed format.

Input

Output - Early

Output - Now

Example 1

Input
Output - Early
Output - Now

Example 2

Input
Output - Early
Output - Now

Example 3

Input
Output - Early
Output - Now

Example 4

Input
Output - Early
Output - Now

Example 5

Input
Output - Early
Output - Now

Transfer

Below is a comparison of voice transfer quality over the past several months. Here the AI takes an original utterance (input) and then transfers that utterance to the voice of a new (target) speaker.

Input

Sample of Target

Output - Early

Output - Now

Example 1

Input
Sample of Target
Output - Early
Output - Now

Example 2

Input
Sample of Target
Output - Early
Output - Now

Example 3

Input
Sample of Target
Output - Early
Output - Now

Example 4

Input
Sample of Target
Output - Early
Output - Now

Example 5

Input
Sample of Target
Output - Early
Output - Now

Denoising

Below are some of our most recent de-noising examples. Some are pure reconstructions, some have transfer as well as denoising.

Noisy Input

Output

Example 1

Noisy Input
Output

Example 2

Noisy Input
Output

Example 3

Noisy Input
Output

Example 4

Noisy Input
Output

Example 5

Noisy Input
Output

Example 6

Noisy Input
Output

Noisy Input

Sample of Target

Output

Example 7

Noisy Input
Sample of Target
Output

Example 8

Noisy Input
Sample of Target
Output

Example 9

Noisy Input
Sample of Target
Output

Example 10

Noisy Input
Sample of Target
Output

Copyright © 2019 VAUX - All Rights Reserved