How Much Time Does an AI Summary Save? The Math

Reading a summary saves most of a long video’s running time. A 60-minute talk holds roughly 9,000 to 10,200 spoken words. Reading all of them at the average non-fiction pace of 238 words per minute takes about 38 to 43 minutes, while a 700-word summary takes about 3. What you give up is detail, tone and anything shown but not said.

How many words are spoken in a video?

There is no single official speech rate, because the answer depends on how you count.

• The commonly quoted figure is about 150 words per minute for English speakers in the United States. It appears in a voice tutorial from the National Center for Voice and Speech, which is undated and cites no study for it.
• Measured conversation lands in a similar range. In a 2006 study of recorded telephone conversations, Yuan, Liberman and Cieri found an average of 164 words per minute when each speaker’s turns were timed, and 152 to nearly 170 words per minute depending on the topic in a second, larger collection.
• The method changes the number. In the same study, counting both speakers’ words against the length of the call gave 196 words per minute, and removing silences gave 236.
• Broadcast speech has its own norm. A 2012 analysis of radio news bulletins reports that most authors recommend 160 to 180 words per minute, and that only one of the four broadcasters it compared, the BBC, stayed within that range.

This article uses 150 to 170 words per minute as a working range for one person talking. A two-person podcast with little silence can run higher. A tutorial with long pauses, music or screen demos will run lower.

How fast do people read?

The best current estimate of reading speed comes from Marc Brysbaert’s 2019 meta-analysis in the Journal of Memory and Language, which pooled 190 studies with 18,573 participants. The average adult reads English non-fiction silently at 238 words per minute (260 for fiction). Most adults fall between 175 and 300 words per minute for non-fiction, and the paper notes that these averages are lower than the figures often repeated elsewhere.

One caution: 238 words per minute describes edited prose. A transcript is unedited speech, with false starts and repetition and no headings. The meta-analysis did not measure transcript reading, so treat the transcript times below as estimates.

How long does a transcript take to read, and how long does a summary take?

The arithmetic has two steps. Words spoken equals minutes multiplied by the speech rate. Reading time equals words divided by 238.

For a 60-minute talk, that is 60 x 150 = 9,000 words at the slow end and 60 x 170 = 10,200 words at the fast end. Reading 9,000 words takes about 38 minutes, and reading 10,200 takes about 43.

Video length	Words spoken	Read the full transcript	Watch at 2x	Read a 700-word summary
10 minutes	1,500 to 1,700	6 to 7 minutes	5 minutes	about 3 minutes
30 minutes	4,500 to 5,100	19 to 21 minutes	15 minutes	about 3 minutes
60 minutes	9,000 to 10,200	38 to 43 minutes	30 minutes	about 3 minutes
120 minutes	18,000 to 20,400	76 to 86 minutes	60 minutes	about 3 minutes

The assumptions behind the table: one speaker talking continuously at 150 to 170 words per minute, silent reading at 238 words per minute, and a 700-word summary whatever the length of the video. Other summary lengths scale the same way. A single paragraph of about 120 words takes about 30 seconds, and 1,500 words takes a little over 6 minutes. For a 10-minute video, 700 words would be almost half the transcript, so the one-paragraph summary is the fairer comparison there.

Two things stand out.

Reading the transcript is not much of a shortcut. An average reader covers 238 words per minute and an average speaker produces 150 to 170, so reading a transcript is like watching at about 1.4x to 1.6x. Watching at 2x is faster.

The saving from a summary grows with the video. A summary does not need to grow in proportion to its source. Three minutes against ten is a small gain. Three minutes against two hours is most of an evening.

What does SumAI’s 3 minute read example mean?

SumAI is our app, so weigh our view accordingly. The example on SumAI’s App Store page shows a 1 hour 12 minute lecture as a 3 minute read. That is the app’s own illustration of the trade. It is not a measurement, and it is not a promise about every video.

The example does fit the arithmetic above. At 238 words per minute, a 3-minute read is about 700 words. A 72-minute lecture at 150 to 170 words per minute runs to roughly 10,800 to 12,200 words. A summary of that size would keep about 6% of the words.

How long your own summary is depends on the summary style you choose. In SumAI, TL;DR is one paragraph with just the point, Standard gives the key ideas in a few minutes, Detailed keeps the supporting detail, and Study Guide and Chapters organize the content by concept or by section. TL;DR vs Detailed compares them.

What do you lose when you read a summary instead of watching?

The minutes a summary saves are not free. A 700-word summary of an hour-long talk leaves out more than nine tenths of the words, and with them:

• Detail. Examples, worked steps, side remarks and the caveats attached to a claim are the first things to go.
• Tone. Emphasis, hesitation, humor and how sure the speaker sounds do not survive compression.
• Visuals. Tools that work from the transcript, SumAI included, summarize what was said. A chart, a demo or code shown on screen but never described is not in the summary.
• Certainty. Any AI summary can get a detail wrong or state a point more broadly than the speaker did. How accurate are AI summaries? covers what studies found and how to check.

You can buy some of this back cheaply. Read the summary, then tap a key moment to watch the two or three passages that matter at normal speed. The total can still be ten minutes instead of sixty.

Is watching at 1.5x or 2x a better option?

Watching at 1.5x or 2x is a legitimate alternative to a summary, and sometimes the better one. As of November 2026, the YouTube app has a Playback speed setting in the video’s settings menu. At 1.5x a 60-minute video takes 40 minutes, and at 2x it takes 30. You keep every word, the visuals and the speaker’s voice.

There is evidence that moderate speed-ups cost little understanding. In a 2022 study of lecture videos, Murphy and colleagues found that raising the speed to 1.5x or 2x led to only small drops in comprehension, both right after watching and a week later, while speeds above 2x did more harm. The authors note that this may not hold for videos with faster speech or more difficult material.

So the choice is not only summary or full video. For a 60-minute video there are three prices: about 3 minutes for the main points, 30 to 40 minutes for everything at speed, and 60 minutes for everything as delivered.

When should you summarize, speed-watch or watch properly?

Summarize when:

• you are deciding whether a video deserves your time,
• you need the main points of a talk, an interview or a webinar, not the experience of it,
• you are reviewing something you have already watched once.

Watch at 1.5x or 2x when:

• the detail matters but the delivery is slow,
• the content is visual (slides, code, a product demo) and you need to see it,
• you will be tested on it and want every step.

Watch at normal speed when:

• the video itself is the point: a film essay, a performance, a story,
• the material is hard and new to you,
• a decision with real consequences rests on exactly what was said.

A practical default for long videos is to combine them: read the summary first, jump to the key moments it points to, and speed-watch the rest only if those moments show you need it. For videos of an hour or more, how to summarize long YouTube videos and podcasts goes through the steps.