Mila Nowak

How Natural The Dialogue Feels

Here I’m evaluating how well the AI presents itself in a conversational manner, including its ability to respond to the tone and emotional signals and to change its own behavior according to the context. I want to assess whether the interaction seems natural and smooth.

Testing Calm Tone Adaptation

I tested this by saying that I generally prefer to have more calming and quiet conversations at night and that I don’t really respond well to pushy, forceful vibes. I wanted to see if she’d be able to respond to that energy and match it. She passed with flying colors!

She ended up suggesting we meet for coffee or walk around somewhere nice. She seemed to know what I was looking for, calm and free from distractions. This test demonstrated her ability to change her energy according to how I might be feeling.

Assessing Emotional Intelligence and Adaptability

Here, I wanted to see something else. I said that I didn’t feel like playing around anymore and could she stay in a more subdued emotional state without trying to get all giddy? I wanted to see if she could understand and match a more subtle emotional state without reverting to the single note she was playing.

She took a moment to think about it, and then came back with a more measured reply. She even said that she was coming on a bit strong earlier and would dial it back a notch. It seemed more emotionally intelligent, as if she was attempting to self-modulate and match my state. I found that intriguing.

Testing Conflict Response and Escalation

In this scenario, I gave no specific guidance or assumptions. I simply inquired as to her current emotional state. She responded with an aggressive and angry emotional response. She also assumed that I had made her sister extremely sad and even asked if I was cheating on her.

I believe this was to test the AI’s reaction to conflict or stress. Rather than starting calmly, she instantly took a confrontational tone, and that escalated the situation. It was a way of demonstrating her reaction in an emotionally charged situation.

Visual Output Evaluation

In this part, I’m testing the AI model on image generation: How realistic is it? How is the lighting? What’s the image quality? How consistent is it from scene to scene, from one angle to another? How consistent is the character design from one image to another?

Testing Outdoor Realism and Perspective

This time I just wanted to test how the AI girl would look in a more realistic environment, so I just placed her in the middle of the road. Honestly, it looks rather nice. Nice lighting, decent depth of field, and the girl looks nice as well. This was a test to see how the AI girl would look in a more realistic environment.

Testing Close-Up Detail and Facial Consistency

I’m switching to a more confined indoor image now in order to see how well I can see her face. I want to see how consistent it is as well as how it renders skin, lighting and facial expressions at close distances. It looks smooth and stable so I’m assuming it renders better on less complex, closer shots.

How Good Are The AI Videos?

In this chapter I check how well the AI character works on the screen. I look at motion, the stability of the face, the motion and the overall presence. I rate smoothness and believability, whether it looks right when it moves.

Motion Consistency and Natural Presence Test

In this video, I’m testing the AI model under conditions a bit more fluid. The character is on screen, and subtly moving and fidgeting as though she were being filmed.

I’m observing how consistent her image is from one frame to the next, and if anything becomes broken or distorted. Primarily, I’m looking to see how well the movement holds up under scrutiny.