Anime Model
Firstly, I began to research how to make an AI assistant
Through research, I found out that I would need to use a programing software called python to create an AI. I also realized that there were only tutorials on making AI talking bots which was not what I was trying to create. Then I continued searching and found a few programs that could demonstrate my prototype: V magic mirror, V see face, prprlive and Vtube studio; they are similar to the software used by Vtubers. Through testing, Vtube Studio turned out to have the best performance overall; basically, Vtube Studio is a Live2D broadcast assistant software that allows Vtubers or anyone to use anime models to replace their physical faces, body and hands.
.png)
Voice lines
After figuring out the anime model, I searched the web for better quality voice lines that are slightly better than Google Translate. Then I found Microsoft's Azure text to speech online service, this was on a website, and the quality of the converter was much more realistic than google translater. Before writting down the scripts for my prototype I asked my client what questions and problems she had that she would like to ask the AI; then I wrote down the AI's script and converted the text and recorded it to create the voice lines.
.png)
Animation
To make the tutorial, I searched youtube for videos on how real people do squats properly. As I watched the video I took 98 pictures of the person doing the squat (to capture and study how muscles move to make a more natural animation) and narrowed it down to 10 photos that were best. This was my first time using Krita, so before animating searched youtube for tutorials to get me started. In the process of using Krita I realised how time-consuming 2D animating was; I had to draw out the outline of every frame with very thin brushes; I was going to colour every frame, but I didn't have enough time, so I at least tried to shade every frame to make it look a little more appealing.
.png)
.png)
After combining footage with Davinci Resolve 17 my prototype was complete.
Feedback from client
I interviewed my client and noted a few issues and suggestions that could improve my prototype.
Do you think anything could be improved?
Do you think some of the information is too much or too less?
Do you think this idea will be better as pc or mobile apps or built into a robot which only are available in the gym?
Do you have anyother suggestions?
I think this idea will work well.
I think the part where Lareina talks about the facts could be reduced somehow, maybe she could answer the questions more straightforwardly (too much informatic info)
Apps will be better than gym robots as it is more portable and maybe you want to use the AI gym assistant at home.
Maybe colour the squat tutorial animation
Normal speed prototype watch or download here
DOWNLOADReflection
From my feedback, I realised that I had too much factual information in my draft prototype, so I deleted some unnecessary information. The additional fact information will be presented if clients ask for it; the AI will only answer the questions without any other irrelevant information.In the first version of the prototype, my AI was one part of a robot that was only accessible inside a gym; from the feedback, I discovered that it would be better if it was an app on Ipads, mobiles or pc that could be used at home as some clients suggested that they may want to work out at home sometimes.The client also suggested that the tutorial animation can be visually improved with shades or colours. Since I ran out of time doing the animation, I did my best to add shades to every frame. If I were to develop this project further, I would make the time-consuming features first, then enhance the project visually and add valuable features.
