Same as when you speak. Speech is a variation in frequency, how loud you speak would be a variation in amplitude. Modern micro electronics can do the same trick with an input voltage.
SFCs are employed when the demands of the machine are asynchronous with the characteristics of the supplied mains either in frequency or voltage. An example in reverse would be a wind turbine which is designed & set up to run at an optimum rpm based on average predicted wind-speed. But seeing as the wind speed is so variable so is the rotational speed of the blades and therefore the output from the genny would vary wildly with windspeed. This is no good for direct (synchronous) connection to the grid as the output frequency of such a 'directly coupled' generator would be unsuitable for connection no matter how good its physical regulation. At best all you could do would be to disconnect it every time the frequency fell out of step and so it would only be contributing power when the windspeed allowed it perfectly to match that of the mains. maybe for 2% of its operational life. That is a huge waste of the total amount of kinetic energy that hits its blades.
The easiest way to recover more energy is to take whatever the generator gives and convert it do DC.
Once you have that regular DC it's a whole lot easier to use a micro-processor controller driving switching thyristors/amplifiers/IGBTs or whatever to create pretty much any wave shape and frequency you want. Even the little micro-processors I fiddle about with run at 4Mhz so they are easily capable of reproducing a perfectly regulated 50Hz (or 400Hz for that matter) sine wave.
