Most people silently pronounce words as they read. It is called subvocalization, and it is not the flaw it is often made out to be.
Subvocalization is the inner speech that accompanies reading — sometimes felt as a voice, sometimes as faint movement in the throat. It is close to universal among fluent readers.
The speed reading argument is straightforward: if you pronounce every word internally, your reading speed is capped near speaking speed. The trouble is that inner speech appears to do useful work, particularly in holding a sentence together while you work out its structure, and in poetry and rhetoric where sound is part of the meaning.
In practice, fluent readers already subvocalize less on easy material and more on hard material. That adjustment is a feature, not a bad habit.
At higher speeds, RSVP naturally reduces subvocalization, simply because words arrive faster than you could say them. Many readers find this comfortable for straightforward prose and uncomfortable for anything they want to hear — a poem, a speech, a passage of Dickens with a long rhythmic sentence.
This is exactly why Wordema does not fix the pace. Slow it down and the inner voice comes back; raise it and the text becomes more visual. Both are legitimate ways to read. Poetry is worth reading slowly.