Paste a link to see how many words a video actually contains, how fast the speaker talks, and how long the same content takes to read instead of watch. No signup for your first few counts.
3 free extractions per day, no signup needed. Counting shares the plain transcript allowance — the counting itself happens in your browser and costs nothing extra.
The count is not an estimate from the runtime. It comes from the video's own caption track — the same one YouTube shows in the player.
youtu.be short link, a Shorts link, an embed link or the bare 11-character video ID all work.Because the transcript is already there, you can reveal the full text and copy it with one button — handy when the word count was really a step towards drafting something from the video.
This is the number most people came for, and the arithmetic behind it is worth seeing plainly. Speech and reading run at different speeds, and the gap is large.
| Activity | Typical rate | Time for 6,000 words |
|---|---|---|
| Conversational speech | 130–160 wpm | About 40 minutes |
| Fast presenter or podcast | 170–200 wpm | About 32 minutes |
| Silent reading, careful | 200 wpm | 30 minutes |
| Silent reading, brisk | 250 wpm | 24 minutes |
| Skimming for one fact | Much faster, with Ctrl+F | Seconds |
The tool uses 200 and 250 words per minute for the two reading estimates, and compares the conservative one against the runtime. Those are averages for ordinary prose — technical material, an unfamiliar language or a transcript with no punctuation all read slower, so treat the figure as a ballpark rather than a promise.
One caveat on the speaking rate: it divides the words by the whole runtime, silences included. A video with a long musical intro, gaps for demonstrations or a silent screen-recording stretch will report a lower words-per-minute figure than the speaker is actually producing while talking.
Anonymous visitors get three transcript extractions a day, shared with the transcript generator, because each one is a real fetch against YouTube. Counting the words afterwards is free — it happens in your browser and costs nothing on our side.
The summariseris the exception, and it draws on a tighter allowance of two AI actions a day. Not arbitrarily: summarising spends model tokens on top of the caption fetch, and plain extraction does not. We'd rather explain that than have you meet it as a surprise error. The counter above the tool always shows the balance the API reports, in its own wording, so it is the real number.
A free account raises both, with credits included and no card asked for.
Every figure on this page inherits the accuracy of the caption track underneath it. For a word count that is mostly fine — a few misheard words barely move a total of several thousand. It matters more if you are working from the text itself.
When the creator uploaded their own captions, the text is punctuated and the words are right, so both the count and the reading estimate are closer to the truth. There is no way to tell from the outside which kind a video has — the quality of the text you reveal is usually the giveaway.
For many videos at once, the same counts come back from the API for a video, a playlist or a whole channel — see the API reference or the cookbook for working examples.
It depends almost entirely on how fast the speaker talks. Ordinary conversational speech runs around 130 to 160 words a minute, so a ten-minute video is usually somewhere between 1,300 and 1,600 words. Fast presenters clear 190. Paste a link above and this tool gives you the exact count from the video's own caption track rather than an estimate.
Words in the transcript divided by the video's duration in seconds, multiplied by sixty. Because it uses the full duration, long silences, music beds and intro sequences pull the number down — a video with a two-minute musical opening will read slower than the person actually speaks.
Usually about a third to a half of the runtime. Most adults read prose at roughly 200 to 250 words a minute while most people speak at 130 to 160, so the transcript of a 30-minute talk is typically 10 to 20 minutes of reading. The tool shows both speeds and the exact time you would save on the video you paste.
Close, but not always identical. The count comes from the caption track, which excludes anything only shown on screen and may render numbers, hyphenated words and filler differently from how you would type them. Auto-generated captions also drop the occasional word in crosstalk. Treat it as accurate to within a percent or two, not to the word.
Not from this page — it is rate limited per visitor and works one video at a time. The API returns the same transcript with its word count and duration for a video, a playlist or an entire channel, so a bulk comparison is a short script rather than a lot of pasting.