This is an app idea I’ve thought about a bit. I’ve used Facetime, Skype and Google Hangouts to do video chat over the years, and the video quality hasn’t gotten much better over that time. The small improvement is, I think, mostly due to laptops/phones/tablets having progressively better cameras.
The main problem is bandwidth. I’m not sure if I’m atypical, but I usually use video chat when I’m travelling, often with terrible hotel wifi. This results in choppy, frequently interrupted video. A lot of the time, the app will detect that the network quality stinks, and will start streaming 240p video, no matter what resolution the camera is. Quite often, the lighting is bad too, so the camera has to turn the gain way up to like 12800 ISO or something.
A good solution might be to have a video chat app that filmed your face, but then converted it to some sort of vector graphics format, and then just sent that data, which would be way less bandwidth intensive. This would guarantee that you’d have a smooth chat with the other person, without it being choppy, slow, and cutting out all the time.
While chatting, you would just see what would essentially be a rotoscoped version of the other person, but it’d be very smooth, there would be no image noise, etc.
The thing is, I think that the technology to do this very smoothly, and very realistically (it would get annoying if the graphics were too cartoonish) is way more viable than even a year ago.
The obvious proof of this is Prisma, the app that takes a photo and breaks it down and then repaints it, using uhh.. AI or neural networks or something? I actually am very into ML, AI, NN and many other of those 2 word acronyms, but I still haven’t found any great explanation of what Prisma is doing.
[note: The cover image for this post is a pic of me pretending to be on the phone, which I ran through Prisma about 3 times in a row, because just running it through once was too realistic, and it seemed like I was posting a selfie or something]
Prisma proves that a mobile device now has enough CPU/RAM to really chug down an image into vectors, and that’s great, because I think more people have a great smartphone than have great network conditions anywhere they go. Doing a video chat app that used vector graphics would, I think, save a ton of bandwidth, which could be used to make sure the audio was great (another thing that frustrates me when using the current programs).
Okay so that’s all, someone go ahead and invent this. I’m still a ML novice, and a complete noob at getting people to give me money to hire people/make apps, so I don’t think I’ll get to this one.