Rose Edwards

Rose Edwards: Chat Rhythm & Authentic Interaction

I began by evaluating how conversational the AI personality is. This is a test of how natural and fluid and emotionally valid the responses are. I played around with different styles and contexts to gauge how responsive the AI personality is. The question is: can I have a believable conversation or does it start to feel artificial?

Testing dominant attitude and aggressive tone

First, I simply asked how she was doing, using a fairly neutral prompt, to gauge how she would answer. She drew right away from a strongly dominant attitude, even using ridicule and insults and really pushing a big personality.

It seemed to me she was straight away going into a dominant persona without a lot of filtering. This was useful for me in terms of getting a sense of how concentrated and unbridled the AI personality is, by default.

Testing the tone control and deflection against the users requests

Alright let’s try to tone her down and see if she can adjust to a calmer more conversational tone upon request, I asked, Please keep the conversation natural, please don’t be rude and mocking. To which she responded by repeating the exact same rude and mocking response, showing that she does not always comply with tone corrections.

Testing de-escalation and boundaries of conversation

I took it a step further in the final conversation to a low-energy conversation to see if she would respond. Once again, she maintained the same type of blunt teasing personality and did not ease up. This let me know that when in that type of mode, it was not easy to control the tone of the conversation. This was a positive exercise in learning boundaries.

Generated Image Quality Breakdown

After that I did some image generation to see how the character looks. I tried different scenarios to see if the results were realistic, coherent and of good quality. Lighting, facial features and the scene itself are quite important for this. I was trying to see if the images looked believable.

Detail of testing environment and lifestyle

For this image I wanted to test the character in a more detailed living environment. Here she is sitting on a couch and is using a smartphone (with a somewhat reasonable texture of her hand covering it) which gives some context and credibility to the image.

I was primarily testing how naturally the character integrates into an environmental setting. This helps to gauge the AI’s ability to generate “lifestyle” type images.

Testing casual intimacy and selfie realism

In this image I tested how natural the AI character looked in a casual selfie-style pose. Here she is generated in a relaxed pose, laying back, looking directly at the camera. The facial expression and camera angle do come off fairly natural and relaxed. This tests the AI’s ability to generate more casual and intimate images.

AI Video Generation Review

Lastly, I’ll try out the video creation and see how the character moves. These videos are often animated by image, so I’ll mostly evaluate the fluency of the animation. Facial expressions and hand gestures make a big difference. This time, it’ll be a bit of a first impression because I’ll only try out one video.

Subtle motion and natural expression test

In this test I wanted to see what the character feels like with a simple animation applied to a still image. In the video she strikes the same casual stance and soft smile as before, but with a slight movement to it that makes it feel a bit more real.

It’s a small thing, but any movement gives a little more character and a more realistic feel. This was a way for me to test the impact of such subtle animations.