A suggested talking speed can nudge silent reading
When a text is framed as something a particular person said, cues about how fast that person talks can modulate how quickly readers move through it. In experiments that varied whether a quoted speaker was implied to be fast or slow, silent reading times shifted accordingly, and the analyses looked at effects at the level of individual word processing rather than only whole passages [1]. The practical picture is that reading speed is not fixed by the words alone; context about who is "speaking" them can tug it up or down.
That matters for the question of whether the effect is limited to hard material. The reading-rate modulation in these quotation experiments was observed in silent reading without a reported restriction to difficult passages [1], which argues against the idea that only demanding texts are sensitive to implied speaking rate.
Text difficulty changes reading speed, but not clearly the size of the cue effect
Difficulty does plainly affect how fast people read silently. One analysis measured silent reading speed on passages at three levels of difficulty and compared those speeds with an average speaking rate of roughly 150 words per minute [2]. That establishes difficulty as a real factor in silent reading speed, but it does not show that the influence of an implied speaking rate appears only when texts are hard.
Put the two strands together and the more defensible answer is that implied speaking rate can modulate silent reading across ordinary word-level processing [1], while text difficulty independently shapes reading speed [2]. The evidence does not isolate an interaction in which difficulty switches the cue effect on or off.
Whether inner speech's articulatory detail carries the effect is untested here
The second half of the question — does the effect depend on how much readers mentally articulate the words — is not answered by these studies. The quotation experiments examined contextual modulation of reading rate in oral and silent reading but did not explore articulatory detail in inner speech [1]. The speed-reading analysis likewise considered speaking rate and text difficulty without examining inner-speech articulation [2].
So a reader can reasonably conclude that implied speaking rate influences silent reading speed and that this is not demonstrably confined to difficult texts, while treating the inner-speech mechanism as a plausible but unverified possibility rather than an established cause.
Sources used in this answer
Contextual modulation of reading rate for direct versus indirect speech quotations
Contextual cues implying a fast or slow speaking rate modulated reading rates in both oral and silent reading, with silent reading analyses examining effects at the level of individual word processing; articulatory detail in inner speech and text difficulty were not explored.
Critical analysis of the study of speed reading
Silent reading speed was measured across passages at three difficulty levels and compared with an average speaking rate of about 150 words per minute; articulatory detail in inner speech was not explored.