Fabian: Difference between revisions

Line 22:

Originally, I wanted to experiment with this new simplicity and develop a narrative around a person who lives "in the screen". I wanted to develop a dialog system that offered choices for questions and answers I defined, which could then be used to converse with the "person in the screen." Depending on the selected question or answer, the "story" would take a different course and hold surprises. Another idea was to place the work in public space. Accordingly, it would not be about speech recognition or generative questions and answers generated by an AI. Instead, I wanted to have a lot of conversations with people here in Weimar, interview them about certain topics, have good talks and then incorporate all of these realistic conversations into the dialog system and record the respective people and their facial expressions with performance capture. In this way, I thought, with enough material, the conversations would be far more realistic and, above all, more personal than those created with well-known chatbots.

But then I looked at some examples on the Internet of MetaHumans, which disappointed me more and more the more I looked into them:

Even though the performance capture results were impressive, unfortunately it was precisely the high realism of the 3D models that overemphasized, to my mind, every little flaw in the facial expressions, speed and movement of the animated heads.

I am sure that this software solution will quickly improve, as it is still in Early Access. For this semester, however, it seems to me that far too much fine-tuning and time is needed to get the movements as realistic as they are shown in the trailer for MetaHuman Creator.

For me it's mainly about the content, the idea of talking to a human being who is convinced that his or her world is in the screen. I don't think it's necessary to have photorealistic faces as interlocutors for that, if the narrative is coherent. While I don't want to rule out the use of MetaHumans at this point, I will definitely focus on a style that tries less to visually mimic reality. Instead, the goal is to create a different reality with its own believable rules.

@@ Line 22: / Line 22: @@
 {{#ev:youtube|watch?v=lDFLrzZy2R4?|640|center|}}
+Originally, I wanted to experiment with this new simplicity and develop a narrative around a person who lives "in the screen". I wanted to develop a dialog system that offered choices for questions and answers I defined, which could then be used to converse with the "person in the screen." Depending on the selected question or answer, the "story" would take a different course and hold surprises. Another idea was to place the work in public space. Accordingly, it would not be about speech recognition or generative questions and answers generated by an AI. Instead, I wanted to have a lot of conversations with people here in Weimar, interview them about certain topics, have good talks and then incorporate all of these realistic conversations into the dialog system and record the respective people and their facial expressions with performance capture. In this way, I thought, with enough material, the conversations would be far more realistic and, above all, more personal than those created with well-known chatbots.
+But then I looked at some examples on the Internet of MetaHumans, which disappointed me more and more the more I looked into them:
+Even though the performance capture results were impressive, unfortunately it was precisely the high realism of the 3D models that overemphasized, to my mind, every little flaw in the facial expressions, speed and movement of the animated heads.
+I am sure that this software solution will quickly improve, as it is still in Early Access. For this semester, however, it seems to me that far too much fine-tuning and time is needed to get the movements as realistic as they are shown in the trailer for MetaHuman Creator.
+For me it's mainly about the content, the idea of talking to a human being who is convinced that his or her world is in the screen. I don't think it's necessary to have photorealistic faces as interlocutors for that, if the narrative is coherent. While I don't want to rule out the use of MetaHumans at this point, I will definitely focus on a style that tries less to visually mimic reality. Instead, the goal is to create a different reality with its own believable rules.