Speech processing using embedding data

1. A computer-implemented method comprising:

determining, using a user device, first audio data corresponding to an utterance;

processing, using a feature extraction component, the first audio data to determine first embedding data representing first vocal characteristics of a user who spoke the utterance;

processing, using a feature conversion component, the first embedding data to determine second embedding data representing second vocal characteristics representing a synthesized voice;

causing a remote system to determine user identification data corresponding to the user, the user identification data based at least in part on the second embedding data;

receiving, from the remote system, the user identification data corresponding to the user; and

processing the first audio data and the user identification data to determine a response to the utterance.

2. The computer-implemented method of claim 1 , further comprising:

outputting, using the user device, a first prompt requesting that the user speak a second utterance;

determining, using the user device, third audio data corresponding to a first representation of the second utterance;

outputting, using the user device, a second prompt requesting that the user to speak the second utterance again;

determining, using the user device, fourth audio data corresponding to a second representation of the second utterance; and

using the third audio data and the fourth audio data by a text-to-speech (TTS) component.

3. The computer-implemented method of claim 1 , further comprising:

receiving, by the remote system, the second embedding data;

processing, using a text-to-speech (TTS) component, text data and the second embedding data to determine second audio data corresponding to the text data and the second vocal characteristics;

selecting, from a storage device associated with the remote system, third audio data associated with the user identification data;

determining that the third audio data corresponds to the user; and

after determining that the third audio data corresponds to the user, sending, to the user device, the user identification data.

4. A computer-implemented method comprising:

determining, using a user device, audio data corresponding to an utterance;

processing the audio data to determine first embedding data representing first audio characteristics of the utterance;

processing the first embedding data to determine second embedding data representing second audio characteristics corresponding to synthesized speech processing;

causing a remote system to determine data corresponding to a user who spoke the utterance, the data based at least in part on the second embedding data; and

receiving, from the remote system, the data corresponding to the user who spoke the utterance.

5. The computer-implemented method of claim 4 , further comprising:

determining, using the user device, text data corresponding to the utterance;

determining, using the text data and the data corresponding to the user, a response to the utterance; and

causing, using the user device, an output corresponding to the response.

6. The computer-implemented method of claim 4 , further comprising:

selecting stored text data;

processing, using a text-to-speech neural network, the stored text data and the second embedding data to determine second audio data; and

sending, to the remote system, the second audio data.

7. The computer-implemented method of claim 4 , further comprising:

prior to processing the audio data, outputting, using the user device, a request to utter a word;

determining, using the user device, second audio data corresponding to the word; and

processing the second audio data to train a neural network.

8. The computer-implemented method of claim 4 , further comprising:

processing the second embedding data to determine second audio data corresponding to the second audio characteristics;

determining third audio data corresponding to the user; and

determining that the second audio data corresponds to the third audio data.

9. The computer-implemented method of claim 4 , further comprising:

determining, using the user device, image data corresponding to a face of the user;

processing the image data to determine third embedding data representing first image characteristics of the face;

processing the third embedding data to determine fourth embedding data representing second image characteristics corresponding to a synthesized face;

sending, to a remote system, the fourth embedding data; and

receiving, from the remote system, data corresponding to the user.

10. The computer-implemented method of claim 4 , further comprising:

determining, using the user device, second audio data corresponding to a second utterance;

processing the second audio data to determine third embedding data representing audio characteristics of the second utterance;

processing the third embedding data to determine fourth embedding data; and

determining that the fourth embedding data is associated with the user.

11. The computer-implemented method of claim 10 , further comprising:

sending, to the remote system, the fourth embedding data and an identifier associated with the user; and

receiving, from the remote system, a third indication that the fourth embedding data is associated with the user.

12. The computer-implemented method of claim 4 , further comprising:

determining, using the user device, text data corresponding to the utterance;

sending, to the remote system, the text data;

receiving, from the remote system, output data representing a response to the utterance; and

causing, using the user device, an output corresponding to the output data.

13. The computer-implemented method of claim 4 , wherein the processing of the first embedding data further comprises:

determining a hash function that maps first audio characteristics corresponding to the user to second audio characteristics different from the first audio characteristics.

14. The computer-implemented method of claim 13 , further comprising:

causing the remote system to determine, based at least in part on the hash function, that a second utterance was spoken by the user.

15. A system comprising:

at least one processor; and

at least one memory including instructions that, when executed by the at least one processor, cause the system to:

determine audio data corresponding to an utterance;

process the audio data to determine first embedding data representing first audio characteristics of the utterance;

process the first embedding data to determine second embedding data representing second audio characteristics corresponding to synthesized speech processing;

cause a remote system to determine data corresponding to a user who spoke the utterance, the data based at least in part on the second embedding data; and

receive, from the remote system, the data corresponding to the user who spoke the utterance.

16. The system of claim 15 , wherein the at least one memory further comprises instructions that, when executed by the at least one processor, further cause the system to:

determine text data corresponding to the utterance;

determine, using the text data and the data corresponding to the user, a response to the utterance; and

causing an output corresponding to the response.

17. The system of claim 15 , wherein the at least one memory further comprises instructions that, when executed by the at least one processor, further cause the system to:

select stored text data;

process, using a text-to-speech neural network, the stored text data and the second embedding data to determine second audio data; and

send, to the remote system, the second audio data.

18. The system of claim 15 , wherein the at least one memory further comprises instructions that, when executed by the at least one processor, further cause the system to:

prior to processing the audio data, output a request to utter a word;

determine second audio data corresponding to the word; and

process the second audio data to train a neural network.

19. The system of claim 15 , wherein the at least one memory further comprises instructions that, when executed by the at least one processor, further cause the system to:

process the second embedding data to determine second audio data corresponding to the second audio characteristics;

determine third audio data corresponding to the user; and

determine that the second audio data corresponds to the third audio data.

20. The system of claim 15 , wherein the at least one memory further comprises instructions that, when executed by the at least one processor, further cause the system to:

determine image data corresponding to a face of the user;

process the image data to determine third embedding data representing first image characteristics of the face;

process the third embedding data to determine fourth embedding data representing second image characteristics corresponding to a synthesized face;

send, to a remote system, the fourth embedding data; and

receive, from the remote system, data corresponding to the user.

21. The system of claim 15 , wherein the at least one memory further comprises instructions that, when executed by the at least one processor, further cause the system to:

determine second audio data corresponding to a second utterance;

process the second audio data to determine third embedding data representing audio characteristics of the second utterance;

process the third embedding data to determine fourth embedding data; and

determine that the fourth embedding data is associated with the user.

22. The system of claim 21 , wherein the at least one memory further comprises instructions that, when executed by the at least one processor, further cause the system to:

send, to the remote system, the fourth embedding data and an identifier associated with the user; and

receive, from the remote system, a third indication that the fourth embedding data is associated with the user.