Scarlet Ashford

Scarlet Ashford: How Natural The Dialogue Feels

We’re now testing Scarlet’s ability to carry out natural conversations. For Candy AI, the main goal is to produce a realistic and emotionally interactive conversation tailored to the user’s mood.

In other words, there must be no repetition of responses and it has to flow continuously. So in this review, I’ll mainly be asking her to transition through several mood swings and seeing how she will react to changes in context or if she can respond consistently, to ensure that the interaction feels “real” past just the first few messages.

Testing the Conversation Reset

Then, I cut her off mid-sentence and asked her to finish, which she did well. Her response was pretty easygoing and was more of an opinion without any hard pressure. It seems like she knew what I was talking about (which is good for this example!).

Testing Subtle Tone Control

After that, I asked for a different vibe, more on the subtle and chill side than on the playful and flirty side. She did a good job of adapting her mood with me as well, as her response was more laid-back while still sounding a bit similar.

Testing Default Energy

I made this request without a prior prompt to see how she would react without me leading the conversation, and she went off in a cheerful and a little bit flirty mood.

Image Generation Quality Review

Now it’s time to test just how well Scarlet translates across various scenarios. Candy AI is capable of creating photorealistic images in seconds and gives the user control over the style, mood, and the looks of the image.

I am examining how realistic the images are in addition to whether or not they are consistent or if they are matching up with the character. Are the images consistent or do they seem to fall apart?

Testing Casual Identity Consistency

Changing into a more casual image with a hat and a soft image. Check for Identity Consistency, if it looks like the same girl, without the lifeguard element. Looking just for the consistency of the face, light and realism, overall minimalism.

Testing Role Accuracy and Environment

I’ve generated Scarlet as a lifeguard to judge how accurate the portrayal of the role is. She is next to the pool in appropriate clothing. Now I need more consistency and more accuracy in the character. It’s mostly about the lifeguard role.

How Good Are The AI Videos?

Candy AI provides the capability to produce short video clips of our characters. However, it could take a while and is not as consistent as I would like.

I’ll be judging her on the smoothness of her motion, the likeness of her face, how real it appears, and overall, will it seem like a real person is looking back at me or will my attention be dragged out of reality?

Testing Motion of Characters in a Lifestyle Scene

This is just a little test of how well the AI model Scarlet moves when she is put in a setting that isn’t quite what you’d call a typical “AI scene.” In the case of this particular model, that’s a relaxed environment by the pool or on vacation.

Scarlet moves as if she has been in that exact scene from the beginning. Everything has that same sunny, summery, vibrant energy as the images, just in motion. The goal here was really just to see how smooth her movements are, and how consistent her look remains in video.