Talking head is the format almost everyone uses, and it is not the one that travels furthest
How speech-led Reels compare with silent ones across 2,157 of them, and one worked example that reached 271 times its following. 21 September 2026.
What the numbers say
Across 2,157 reels, 1,890 are speech led and 267 are not. The speech led ones run a median of 51.3 seconds and carry 160 spoken words. The silent ones run 15.9 seconds. On raw popularity the talking heads look dominant, but measured against the audience behind them they reach a median of 0.50 views per follower against 0.63 for the silent ones, and 9 percent of them clear ten times their following against 15 percent.
Two things follow, and the second is the useful one. First, if you are making talking head videos you are in the large majority, so the format itself is not what distinguishes you. Second, the gap is small and the silent group is short: 15.9 seconds against 51.3. A silent reel gets one idea across and stops, which is a discipline rather than a format. The worked example below is speech led and reached 271 times its following from an account of 9,127, so it shows what the majority format looks like when it does work.
Sample of 2,157, measured on 2026-09-21. Scrptly production database, queried 2026-09-21. Instagram Reels only, deduplicated by shortcode, one query run.
The measurement
Every deduplicated reel with a view count, a follower count and a duration, split by whether its transcript holds real spoken words or a trending audio placeholder. The two groups hold 1,890 and 267 reels and add to 2,157. Views per follower and length are medians within the group.
| Measure | Speech led | No speech |
|---|---|---|
| Share of all reels | 88% | 12% |
| Median views per follower | 0.50 | 0.63 |
| Median length | 51.3 seconds | 15.9 seconds |
| Median spoken words | 160 | 2 |
| Reached ten times their following | 9% | 15% |
What that looks like end to end
Input: https://www.instagram.com/reel/DYKp_rcPTqz/ pasted into Scrptly, no other instruction given. A speech-led reel from an account with 9,127 followers.
73.8 seconds, 179 spoken words, which is a little over the 160 median for a speech-led reel and well inside normal. The opening line is the whole mechanism: what if every man on Earth suddenly disappeared. It is a question, it is about the viewer rather than the creator, and nothing in it tells you who is speaking or why. The reel earned 2,477,185 views against 9,127 followers, which is 271 times the following, where the median speech-led reel in this corpus reaches 0.50 times it.
Score returned: 88. 88 is the retention score our analysis gave it, against a corpus median of 82. The hook scored 10. Both are above average, though the thing that actually separates this reel is the 271 times reach, not the two or three points on either score.
What to do next
- Count the words in your script against the 160 median. A talking head reel past about 51 seconds and 160 words is longer than the median here, which is fine if the idea needs it and expensive if it does not.
- If a reel does not need you on camera, try it silent. That group runs a fifth the length, reaches slightly further per follower, and breaks out more often. The constraint appears to be doing some of the work.
- Judge the opening line on whether it works for somebody who does not follow you. The worked example above never says who is speaking, and it reached 271 times its following from an account of 9,127.