PeCap Sem Break '24: Cleanup and Formative Hand In!

What a week. Or should I say what felt like a month. In its entirety, the cleanup process has taken me seven full days to complete. Part of me expected this (thanks to Dr Kennedy's meticulous schedule of the cleanup), and part of me refused to believe it altogether.

I spent one full day cleaning the brows, two days on the eyes and cheeks, and finally FOUR and a HALF solid days on the mouth. Hence why I'm making another blog post to tell you about that process.

I knew there was going to be a lot of cleanup involved to remove jitter and hand animate aspects such as the eyeballs and tongue movements.

Here is a picture of what my setup looked like as I was analysing the brows. I put the viewport window down so low so that there was minimal distraction from the rest of the face. On another screen I had my graph editor. The brows was by far the easiest part to clean and in hind sight, I think I took far too long being incredibly meticulous with them. I wanted to capture every single jitter to make it as close to the reference as possible.

Then I moved onto the eyes. I did a similar thing with my window so that I could only see the brows and the eyes.

One thing I really enjoyed about the cleanup of the eyes was keyframing the eyeballs. It was really satisfying to see the final result as I got the oppotrunity to nail down the natural jittery movements that weren't captured by Retargeter. Analyser struggled with tracking the eyeballs and this resulted in an extremely smoothed eyeball sequence which just isn't realistic.

The actress Muna was extremely expressive so at times I struggled to capture how wide her eyes opened as the digital model isn't 100% accurate to the human. This caused some quite uncanny looking eyeballs.

In this video you can see how the eyeball on the left (Miri's right) that I cleaned looks compared to the right eyeball.

Onto the mouth. I've never been so overanalytical about someone's mouth in my entire life. The cleanup process gave me the perfect opportunity to strengthen some of the posing I did which I was unhappy with at the end of Week 7.

One thing that I found incredibly useful that Dr Kennedy showed to us was buffer curves. Not only was it satisfying to see the raw retarget graph compared to the cleaned graph, but it also helped me understand what retargeter was trying to achieve when I moved a key, and why. I could go back and forth comparing my new key location to the original key location. Also, Retargeter had a lovely habit of overshooting, making gradual increase/decreases for no reason, and spiking in the complete opposite direction before making a large movement.

I will admit, I think I overused my animators brain to smooth out the curves - using ease ins and outs especially.

But one thing that stuck with me throughout this whole process is a quote from our Week 6 lecture. It featured a video on tackling the uncanny valley in which Will Jackson, founder and CEO of Engineered Arts, talked about capturing the 'essence of being human' through not just the expressions themselves but the transitions between expressions. Because of the fluidity of the human skin, I felt that ease ins and outs helped make the expressions transition more subtly.

Finally after my long journey cleaning up the mouth, I worked on animating the tongue. This process was a lot shorter than I anticipated as she keeps her mouth closed (and therefore her tongue hidden) for quite a few frames.

Now I feel like I've yapped enough about my thoughts and feelings on the cleanup.

This is the result of my week and a half long efforts. It isn't perfect for sure but I just can't look at her for any longer at this point or my brain might explode.

Here is my final formative hand in (with the original audio).

Some things I'd like to take note of:

• In the 'cleanup DONE' video, there is considerably less movement on her face in the frames before she smiles compared to the reference video. When I was cleaning it up, I took note of the fact that this apparent jitter was moreso caused by the motion of her head nodding than a change in her expression.

• In the 'Formative' hand in video, the camera has been shifted to the left so the perspective has changed. I think this caused her left (our right) mouth corner to be significantly lower than the other corner. This might have been because the reference video camera was slightly perpendicular to her face angle and this resulted in one side of her face appearing higher than the other. I'd like to believe that the accuracy of the performance is enough to make up for the tilt of her lips.

Tune in next semester for work on the body (yippee.).