System and method for generation of human like video response for user queries

1. A method for generating video response for user queries, the method comprising:

receiving, by a video generator, a visual image of a character of interest from the user;

generating, by the video generator, a synthesized frontal face of the character of interest;

generating, by an encoder-decoder logic in the video generator, an audio sequence and a video sequence for one or more predetermined textual responses, wherein the audio sequence and the video sequence are generated by:

generating a symbolic-token-sequence associated with the one or more predetermined textual responses by mapping an associated symbol to each textual input included in the one or more predetermined textual responses;

generating a numerical vector representation of the one or more predetermined textual responses by embedding the symbolic-token-sequence in a real-valued vector space, in a character-by-character scanning manner; and

generating a plurality of video frames with corresponding internal states in a time sequence, by processing the numerical vector representation of the one or more predetermined textual responses and the synthesized-front-face based on an interactive framework,

wherein the interactive framework enables generation of each video frame with respect to a current-state in time depending on a current-state textual input and on all video frames generated at previous-states in time, and

wherein the plurality of video frame include the synthesized front face with one or more facial expressions that matches to the textual inputs;

generating an audio sequence corresponding to the one or predetermined responses using a Wave Net architecture; and

generating, by the video generator, the video response by combining the video sequence and the audio sequence, for the one or more predetermined textual responses.

2. The method as claimed in claim 1 , wherein the one or more facial expressions of the character of interest comprise lip movement and eye movement of the character of interest, wherein the lip movement matches pronunciation of the one or more predetermined textual responses.

3. The method as claimed in claim 1 further comprises determining gender of the character of interest based on the visual image.

4. The method as claimed in claim 1 further comprises modulating vocal rhythm of the audio sequence based on gender of the character of interest.

5. The method as claimed in claim 1 , wherein combining the audio sequence and the video sequence further comprises synchronizing the audio sequence with the one or more facial expressions of the character of interest.

6. The method as claimed in claim 1 , wherein the encoder-decoder logic is trained using the interacting framework of: Textual Gated Recurrent Unit (TGRU) based Recurrent Neural Network (RNN) video encoder along with following logics: Convolutional Neural Network (CNN) image encoder, and a Conditional Pixel CNN (CPCNN) decoder using training data.

7. The method as claimed in claim 1 , wherein the encoder-decoder logic is trained using the interacting framework of: a convolutional Long Short-Term Memory (LSTM) based RNN video encoder along with following logics: Convolutional Neural Network (CNN) image encoder, and a Conditional Pixel CNN (CPCNN) decoder using training data.

8. A video generator for generating video response for user queries, the video generator comprising:

a processor; and

a memory communicatively coupled to the processor, wherein the memory stores processor-executable instructions, which, on execution, causes the processor to:

receive a visual image of a character of interest from the user;

generate a synthesized frontal face of the character of interest;

generate an audio sequence and a video sequence for one or more predetermined textual responses wherein the audio sequence and the video sequence are generated by an encoder-decoder logic in the video generator by:

generating asymbolic-token-sequence associated with the one or more predetermined textual responses by mapping an associated symbol to each textual input included in the one or more predetermined textual responses;

generating a numerical vector representation of the one or more predetermined textual responses by embedding the symbolic-token-sequence in a real-valued vector space, in a character-by-character scanning manner; and

generating a plurality of video frames with corresponding internal states in a time sequence, by processing the numerical vector representation of the one or more predetermined textual responses and the synthesized-frontal-face based on an interactive framework,

wherein the interactive framework enables generation of each video frame with respect to a current-state in time is generated depending on a current-state textual input and all video frames generated at previous -states in time, and

wherein the plurality of video frame include the synthesized front face with one or more facial expressions that matches to the textual inputs;

generate an audio, sequence corresponding to the one or predetermined responses using a Wave Net architecture; and

generate the video response by combining the video sequence and the audio sequence, for the one or more predetermined textual responses.

9. The video generator as claimed in claim 8 , wherein the one or more facial expressions of the character of interest comprise lip movement and eye movement of the character of interest, wherein the lip movement matches pronunciation of the one or more predetermined textual responses.

10. The video generator as claimed in 8 , wherein the processor is further configured to determine gender of the character of interest based on the visual image.

11. The video generator as claimed in claim 8 , wherein the processor is further configured to modulate vocal rhythm of the audio sequence based on gender of the character of interest.

12. The video generator as claimed in claim 8 , wherein to combine the audio sequence and the video sequence, the processor is further configured to synchronize the audio sequence with the one or more facial expressions of the character of interest.

13. A non-transitory computer readable medium including instructions stored thereon that when processed by at least one processor cause a video generator to perform operations comprising:

receiving a visual image of a character of interest from the user;

generating a synthesized frontal face of the character of interest;

generating an audio sequence and a video sequence for one or more predetermined textual responses wherein the audio sequence and the video sequence are generated by an encoder-decoder logic by:

generating a symbolic-token-sequence associated with the one or more predetermined textual responses by mapping an associated symbol to each textual input included in the one or more predetermined textual response;

generating a numerical vector representation of the one or more predetermined textual responses by embedding the symbolic-token-sequence in a real-valued vector space, in a character-by-character scanning manner; and

generating a plurality of video frames with corresponding internal states in a time sequence, by processing the numerical vector representation of the one or more predetermined textual responses and the synthesized-frontal-face based on an interactive framework,

wherein the interactive framework enables generation of each video frame with respect to a current-state in time depending on a current-state textual input and on all video frames generated at previous-states in time, and

wherein each video frame includes the synthesized front face with one or more facial expressions synchronized to the textual inputs;

generating an audio sequence corresponding to the one or predetermined responses using a Wave Net architecture, and generating the video response by combining the video sequence and the audio sequence, for the one or more predetermined textual responses.