Word replacement during poor network connectivity or network congestion

1. A method, comprising:

obtaining a first audio stream from a user device connected to a real-time communication session;

detecting first speech data in the first audio stream;

converting the first speech data to first text data including one or more words;

determining that the first text data has a missing word based on a context of the one or more words;

synthesizing a predicted word for replacing the missing word in a voice of a user of the user device to obtain a synthesized word;

combining the synthesized word with the first audio stream to generate a second audio stream, wherein the first audio stream is stored in a buffer for a predetermined duration to combine the synthesized word with the first audio stream; and

transmitting the second audio stream to one or more other user devices connected to the real-time communication session.

2. The method of claim 1 , further comprising:

transmitting a notification to the one or more other user devices that indicates that a portion of the second audio stream is synthesized.

3. The method of claim 1 , wherein the second audio stream contains second speech data, the method further comprising:

converting the second speech data to second text data; and

transmitting the second text data to the one or more other user devices.

4. The method of claim 1 , wherein the first audio stream is stored in a buffer to combine the synthesized word with the first audio stream.

5. The method of claim 1 , further comprising:

determining a speech profile of the user based on real-time communication session recordings associated with the user; and

determining the missing word based on the speech profile.

6. The method of claim 1 , further comprising:

determining the context of the one or more words based on a neighboring word range for a keyword of the one or more words.

7. The method of claim 1 , further comprising:

transmitting a request to the user device; and

receiving, in response to the request, a message that indicates whether the synthesized word is accurate.

8. A system, comprising:

a first user device connected to a real-time communication session;

and

a server configured to:

obtain a first audio stream from a user device connected to the real-time communication session;

detect first speech data in the first audio stream;

convert the first speech data to first text data including one or more words;

determine that the first text data has a missing word based on a context of the one or more words;

synthesize a predicted word for replacing the missing word in a voice of a user of the user device to obtain a synthesized word;

combine the synthesized word with the first audio stream to generate a second audio stream, wherein the first audio stream is stored in a buffer for a predetermined duration to combine the synthesized word with the first audio stream; and

transmit the second audio stream to one or more other user devices connected to the real-time communication session.

9. The system of claim 8 , wherein the server is further configured to:

transmit a notification to the one or more other user devices that indicates that a portion of the second audio stream is synthesized.

10. The system of claim 8 , wherein the server further configured to synthesize the one or more predicted words using a deep neural network.

11. The system of claim 8 , wherein the missing word is determined based on a speech profile.

12. The system of claim 8 , wherein the server is further configured to:

determine a speech profile of the user based on real-time communication session recordings associated with the user.

13. The system of claim 8 , wherein the server is further configured to:

determine the context of the one or more words based on a speech profile of the user.

14. The system of claim 8 , wherein the server is further configured to:

transmit a feedback request to the user device; and

receive, in response to the feedback request, a message that indicates whether the synthesized word is accurate.

15. A non-transitory computer-readable medium comprising instructions stored in a memory, that when executed by a processor, cause the processor to perform operations comprising:

obtaining a first audio stream from a user device connected to a real-time communication session;

detecting first speech data in the first audio stream;

converting the first speech data to first text data including one or more words;

determining that the first text data has a missing word based on a context of the one or more words;

synthesizing a predicted word for replacing the missing word in a voice of a user of the user device to obtain a synthesized word;

combining the synthesized word with the first audio stream to generate a second audio stream, wherein the first audio stream is stored in a buffer for a predetermined duration to combine the synthesized word with the first audio stream; and

transmitting the second audio stream to one or more other user devices connected to the real-time communication session.

16. The non-transitory computer-readable medium of claim 15 , the operations further comprising:

transmitting a notification to the one or more other user devices that indicates that at least one word in the second audio stream is synthesized.

17. The non-transitory computer-readable medium of claim 15 , wherein the second audio stream contains second speech data, the operations further comprising:

converting the second speech data to second text data.

18. The non-transitory computer-readable medium of claim 15 , operations further comprising:

determining a speech profile of the user based on real-time communication session recordings; and

determining the missing word based on the speech profile.

19. The non-transitory computer-readable medium of claim 15 , wherein the predicted word is synthesized in the voice of the user using a vocal model.

20. The non-transitory computer-readable medium of claim 15 , the operations further comprising:

receiving a message that indicates whether the synthesized word is accurate.