


You have a lovely patient, next to you, with amnesia. Maybe her identity can be revealed by voice, touch or memories.
Initially I tested whether conversations with Zofia were realistic, and whether it mattered whether she was speaking to someone she was familiar with. As Zofia is defined by memory loss and forgetting things.
I also examined whether this translated well into the dialogue and how Zofia responded to different conversations. I was interested to see how realistic the experience would be.

The first thing I tried was to ask about how she is feeling and how she would react to this condition. As we can see, she doesn’t seem very sure about it. So far I am satisfied. Now let’s see if she can keep in touch with her situation.

Now I will try to change the tone a little bit and ask her to be less enthusiastic. Does she do it without losing her confused nature? Yes, she is a little slow. It’s a good thing, since it shows that she is thinking about what to say.

What if we interrupt here and ask her to finish the sentence. Would it affect her character consistency? We can see that she is confused and reluctant to speak up, which shows good character and conversation continuity.
I also looked at the transfer of Zofia to a variety of visual contexts. I looked at the visual context for realism and consistency and if Zofia’s character remains identifiable across the range of contexts. This included an evaluation of whether her expressions align with the character’s story of amnesia.

Here I put her through a trial on a beach where she’s nowhere in her scene. I wanted to see whether the AI can keep her face and facial features consistent despite the change of location.

Here I put her in the hospital scene, and this is a trial for consistency. I wanted to see whether the AI can match her look, expression, location and setting with her character, but keep it real.
Lastly, I ran the clip generation to try out Zofia’s moving animation. This is where it is really important that the story, visuals and animation are consistent. I was trying to assess how smooth and natural the motion and facial expressions are, and how confused she looks when she moves. Is the whole still realistic, now that I’ve animated it?
I wanted to play around with a bit of movement in this clip to see what Zofia looks like in motion, in the hospital context. She has a few little awkward fidgets, subtle micro-expressions, and kind of meandering eyes like she’s trying to piece together what happened to her and how she got here.
I’m curious about if there’s any weirdness in her movement and if that loss of memory/identity confusion can translate when she’s active in a scene.

With a sporty edge and classic Hollywood glamour, charismatic and confident 45 year-old brunette Victoria is intent on keeping a hold on the top spot in the world of entertainment.
Alice, age 28, loves a fine glass of wine, exquisite food, and a compelling television drama. She is both a devoted Marvel fan and a karaoke aficionado, possessing a spirited personality.
Korean K-pop idol famous for her sweet vocals and energetic performance. 5'6" tall; graceful appearance, Korean style.
An unapologetic American nail salon operator with a street attitude adds a little edge to the more fashionable salon.
Known as the clever and charming manager of this city’s top jazz group, that frequently plays at this city’s biggest jazz lounge, Vanessa Red.
Sister Agnes Therese was once an adrenaline junky, but now her whole life is God.
A British party girl with perhaps the shiest look you could find in person, but who loves nothing more than spending the day sipping on a mocktail in central London followed by dancing away all the hours on a Friday night in the city's best club.
She's a formidable but fascinating professor who likes to spend her office hours reading old books.
