Estimate text-to-speech duration.
Text-to-speech services take character counts as input and bill by them, so the estimate starts there and converts to words at roughly 5.7 characters per word including the trailing space. Synthetic voices default to about 175 words a minute at a rate setting of 1.0, faster than human narration because they need no breath. The rate multiplier scales that linearly. What gets forgotten is explicit pause markup: a 350 millisecond pause inserted at each of 42 sentence breaks adds nearly fifteen seconds, which matters when you are fitting synthesised audio to a fixed video slot.
Word estimate
Words = characters / 5.7 including spaces
Duration
Seconds = words / (base rate x rate multiplier) x 60 + sentences x pause ms / 1000
Because that is the unit TTS services measure and charge in. Working from characters means your duration estimate and your cost estimate use the same number, and it avoids a word-count mismatch when the text contains numerals, abbreviations or markup.
Yes, more than people expect. Explicit pause markup at every sentence break accumulates fast — 350 milliseconds across 42 sentences is nearly fifteen seconds, which can be the difference between fitting a fixed video slot and needing a re-edit.