Question

I am drafting architecture for 1:1 video calling mobile app with face filters (face recognition).

I could use p2p e.g. WebRTC because more savings less complexity. However, continuous face rec during a video call on cheap phones becomes more expensive for the device.

Is this cost increase generally significant enough to warrant routing the traffic through a server?

Was it helpful?

Solution

Sometimes you just need to decide what the minimum hardware requirements are. Mobile devices are a fickle bunch. Small. Not as powerful. Spotty data connections. You could have a valid business reason for choosing client side filtering for video. Then it is a research effort on your part to figure out which mobile devices are in use, what their hardware is, and how much that limits your user base (and therefore potential customers). It's all a balancing act, because doing all this filtering on the server could be expensive too, either in licensing costs or the shear amount of hardware to support the myriad users you (hope to) have.

OTHER TIPS

Decide who is going to pay for it, the user or you. Lots of people have powerful phones. A new iPhone has about as much power as three cores on your server, plus specialised hardware for that kind of task.

Licensed under: CC-BY-SA with attribution
scroll top