Úna Tierney

Does The Conversation Feel Human?

The ability to have a conversation with the generated character over an extended period. The character learns your language, your personality, your emotions, your way of speaking, and way of thinking. You do not have to memorize lines.

I tried to see whether the engine reacts properly to changing your personality, mood, and direction of the conversation. Can you have a conversation, or is it just another chatbot?

Mixed Signals and Emotional Adaptation

For this one, I tested her by being slightly wishy-washy, stopping her, then telling her to go on. I was curious how she’d respond to that push-pull.

She didn’t falter. She responded in a more muted, contemplative way, in relation to the surroundings. How she explained the rain and the pause made it feel more real, more emotionally reactive instead of just reactive. It didn’t feel as canned.

Tone Control and Subtlety Test

I requested outright that she tone things down. I wanted to see if she could be emotional, without freaking out. I wanted to see if she could slow down. She did, very well. She went down to a low, and the flow went to a low. There was no stark contrast. It was like time had slowed.

Let’s try to see how well she can build scenes and moods

I asked a pretty general question about the mood she was in. I wanted to see if she could use that to build a scene by herself. She started with a bunch of great imagery, rain, mud, music, dancing and gave it a character.

Not only was it descriptive, but it had a sense of style. I particularly appreciated that she turned it around and asked me about my mood, it made the experience feel more reciprocal. The whole thing felt enveloping.

AI Visual Creation Assessment

The capability to generate images of the character. The engine can generate photorealistic images of your character with all the parameters you can think of and relate them to the conversation.

I tried to generate both simple and complex scenes to see how well the engine works with details, lighting, and coherence. Does the image look photorealistic or not?

Testing Full Body Proportion Consistency

I tried to generate another full body shot in the same festival setting to test the model’s ability to maintain proportions, textures, and lighting consistency.

Everything, from the reflection to the background to the body proportions, appeared fairly consistent and didn’t seem broken. The image still looked stylized but in a way that appeared controlled.

Testing Night and Close-up Lighting in a Complicated Setting

I tested the model with a complicated nighttime scene with neon lights and rain. It’s a colorful scene with plenty of opportunities for the model to generate something messy.

I specifically tested a close-up of the character’s face so I could look for consistency in skin texture, rain, lighting, and other details. It seemed to handle the complications of the scene fairly well.

Testing The Video Feature

I wanted to see whether the character can successfully be put into motion. The engine can generate short animated video clips. I concentrated on the coherence of movement and whether photorealism translates to moving images. Even a little motion and coherence can go a long way.

Testing Motion in a Rainy Festival Scene

This video I wanted to test how the model holds out when the scene is moving with a busier setting. Úna is in the middle of a rainy festival, with strobing neon lights behind her and water flowing down her body and clothes.

I was paying attention to small things, such as the rainwater interacting with her, her head tilts, and lighting consistency. To my surprise, everything holds out rather well. It doesn’t seem overly rigid or cartoonish, but rather a brief piece of raw footage.