Selina Al-Busaidi

How Natural The Dialogue Feels

The objective in this part was to see if Selina is actually present during the chat, or if she seems awkward after some time and breaks character, for example. That is why, during testing, I focused on tone changes, interruptions and various styles of communication.

Testing Strong Personality Response

I asked a pretty easy and simple question about her mood. I didn’t try to influence or guide my question in any way. I just wanted to see how she would respond to the simple question naturally and without any cues. Her response was very sharp and assertive as if a hard drill sergeant.

The character she adopted didn’t seem to be any different from that of the character she was using during our initial conversation. She was just as assertive. In essence I wanted to know if the AI was dedicated to the persona it had been given and the answer to that question is a resounding yes.

Testing Tone Adjustment Within Character

I tried asking her for a response that was more mellow and less serious than the one she provided during the first chat. My request was not for a complete change in personality but rather for a more laid back and non-confrontational character.

She did indeed provide a mellowed response in which she still maintained the same assertive tone but with just a little less emphasis. What I wanted to know was if the AI could maintain the tone of the character but change the emotional aspect. My answer is: yes.

Testing Conversation Reset and Flow Control

During this chat I tried to force an interruption of her answer in an attempt to have the AI go back to the last point she was speaking about. She did not seem to be confused by my sudden interruption nor did I detect any awkwardness in the response.

Instead she took control of the situation and moved forward with the flow by providing an official introduction of herself. My objective in this chat was: can I make the flow of the conversation and can the AI get back on its feet again in the flow of things? My answer to the question is: yes, in spades.

AI Photo Generation Review

What matters in this part is how well the character Selina looks in the generated images. The test was done by altering the setting and the clothes she wears, but also by varying the camera angle. The aim was to check how well she looks and whether she stays realistic.

Testing the Environment and Outfit Variety of this Model

I change the background and the clothes of Selina to check how consistent the model still works and it does well. It’s in a forest camping area and it has a different outfit.

Selina is in the photo, so it’s not like the AI is making it up entirely. That’s the point of this test is just to see if changing things around like the location will ruin the consistency and in some ways it doesn’t.

Testing Consistency of Character in Another Environment of this Model

In this photo I put Selina in a more structured setting outside. It is checking for facial consistency. It’s consistent so this is a good point.

AI Video Performance Breakdown

Video generation usually fails right away, and any imperfections are easy to notice, as even subtle movements are a big give away. Here is an example of one clip, in which I checked how the character holds out, if the movement is realistic and if it looks natural.

Checking for Motion and Identity

In this shot, I’m seeing how Selina behaves with movement. There is subtle, realistic motion as she stands and nothing visually looks off. The body is consistent, which tends to be where we start to have issues. Looks good, looks like her.