Bullshit, physical impossibility. There are some tricks you can use to jump through material quickly, but they're just tricks and they're prone to error.
Some explanation on why it's bullshit, and what some of these tricks are:
The fovea of your eye, which is the only area that has enough detail perception to be able to read, flicks around 3-4 times per second unconsciously. The fovea will flick to interesting things in the visual space, and your brain "remembers" this detail and stitches in actual observed detail into your perceived vision along with a lot of "fake detail" which is your brain just approximating based on a variety of things I won't get into. But your vision, despite what it seems, is nothing like a camera.
With speed reading, your fovea will flick 3-4 times over the page, in a second, and will home in on particularly attention grabbing words. Your brain may or may not actually get the detail of a "chunk" of text (the fovea is roughly equivalent to the % of your visual field that your thumbnail will occupy at arm's length, e.g. you won't be reading much).
From doing this, you can sort of get the "gist" of what's going on very rapidly. But you're at the mercy of a.) a very, very small bits of disassociated text on the page, b.) that your brain pieces together the information in an appropriate manner, c.) that your focus of attention will be drawn to actual pertinent information. And this isn't really reading, not as the word is usually meant/
There is no way to speed read as depicted in popular media. It's just a flat impossibility that people think can be done because they don't know how their consciousness actually works. At best exists a few tricks to maybe extract the gist of a text at a very fast rate, but it's wildly inaccurate, not in your conscious control, etc.
An alternative method, that I can do, is simply reading something like 3 sentences at once. Which isn't what I'm actually doing, I'm scanning several lines for bits that jump out, ignoring filler or non-meaningful words, and then stitching together meaning based upon what words seem meaningful, where they're located. But this isn't fool proof, as lots of filler words aren't filler words, lots of subtle nuances in a sentence can change its meaning, my brain may misremember or assemble incorrectly the order of the important bits that create an illusion of a different meaning, etc.
edit: I've read 7 (or 8) 500+ books in one day where I had not much to do, and that wasn't the only thing I did that day. I read extremely fast compared to most people. Some people say I'm a speed reader. But that's very, very different than what "speed reading" refers too. Why I read so fast is a.) a lot of practice, b.) the type of intelligence I have. "Speed reading" refers to some trick or gimmick that will suddenly let you read in a different way than other people. I don't read in a different way than most people, I'm just much, much faster than your typical person.
As to subvocalizing (saying the words in your head):
Yes, everyone will start out subvocalizing. Subvocalization engages very specific cognitive machinery (e.g. modularized areas of your brain that activate in regard to different stimuli). That's why little kids move their lips when they learn to read, because they are actually engaging the mental mechanisms involved in speech, and their brain is actually acting as if they were speaking the words that they were reading (down to the nerves that control your vocal chords firing as if you were speaking). Most kids will learn to stop doing this as it's superfluous, alterable by conscious control, and slows them down as actually moving your tongue and lips is a pretty slow process. And most people subvocalize all their lives, it's pretty normal.
You can stop subvocalizing while reading, but this severely hinders your ability to draw meaning, to parse complicated sentences, and severely biases your interpretation of what you are reading in favor of what your brain decides to create from the text. Note, this isn't a matter of conscious interpretation, but of processes beneath the conscious level. Thus, you're not really reading in any real sense, you're creating a story in your head using the text in front of you as a cue. At best you'd be a poor but fast reader, at worst what you think you're reading will have very little bearing on the text in front of you.
A way to bypass some of these problems is by using ideas instead of subvocalization, replace saying "cat" with the idea of a "cat". But this will engage very different machinery in your brain, and alter the way you perceive the text. Some people will never be able to do this, most won't be able to do it well, and those who can do it very well (e.g. high congruency between thought based on subvocalization based reading) will be very, very few. This isn't a matter of how smart you are, but more of where your particular intellectual strengths and weaknesses exist. You'll have to be heavily dominant in verbal areas, not how high you score, but what your comparative scores between two areas are. Being so heavily dominant in one sub-set of intellect is not a good thing. It can qualify as a learning disability and will have pervasive effects. It doesn't mean you're dumb, in fact you can be very, very far from dumb, you can score far above the baseline in every area on a 12 hour test by a psychometrist, but because you're so heavily dominant in one area it can screw you around.
edit: people who scan academic texts are crrrr-aaazy. You're not actually "reading" faster, there's no panacea, you're sacrificing your ability to understand detail for getting to the end faster. If you already thoroughly understand the information, yeah, sure, scan for bits you don't know, but then read them.
Disclaimer: this falls precisely within my area of (near) expertise in academia. If I seem in a rush to throw around assertions, it's because this has been studied heavily. And, in regards to cog psych, people's perceptions on how their minds work, even on seemingly trivial matters of conscious perception, are nearly invariably in disagreement with the body of empirical knowledge and testing.
