Guide
Talking head video statistics: sourced, dated and checked
Faces on camera have a real evidence base, and this page gathers it: peer-reviewed research on faces, retention data from the hosting platforms, caption studies and creator-economy figures. Every statistic names its source and year in the sentence, and the famous numbers that failed the check get their own section instead of another repeat. Last verified: August 2026.
- How fast the eye finds a face (Journal of Vision, 2010)
- 100 ms
- More likes for photos with a face, across 1.1M posts (Georgia Tech and Yahoo Labs, 2014)
- +38%
- Average engagement on videos under a minute (Wistia, 2025)
- 50%
- Numbers on this page without a named source and year
- 0
The perception research: eyes go to faces first
The case for putting a person on camera does not start with marketing surveys. It starts with vision science, and the vision science is one-sided.
For a coach, advisor or consultant deciding between a faceless slideshow and their own face on a weekly take, these are the numbers that settle it.
- 100 to 110 milliseconds: how quickly the human eye can begin a saccade toward a face in a scene, faster than toward any other object tested (Crouzet, Kirchner and Thorpe, Journal of Vision, 2010).
- 38% more likes for photos containing a human face, in an analysis of 1.1 million Instagram photos; the age, gender and number of faces made no difference (Bakhshi, Shamma and Gilbert, Georgia Tech and Yahoo Labs, 2014).
- 32% more comments for those same face-containing photos, from the same 1.1-million-photo dataset (Georgia Tech and Yahoo Labs, 2014).
- 6.9 million video-watching sessions analysed on edX found informal talking-head video more engaging than polished studio productions of the same material (Guo, Kim and Rubin, ACM Learning at Scale, 2014).
- The same study found engagement falls off sharply beyond about 6 minutes, regardless of production quality (Guo, Kim and Rubin, 2014).
- In a classroom study of 200 students across 72 videos, lectures combining a talking head with on-screen annotations produced the longest watch time and the highest satisfaction and engagement scores of the formats tested (Garcia and Yousef, 2023).
Retention by length: what the hosting platforms measure
Video hosts can see what viewers actually do, not what they claim in surveys. Two of them publish it.
For a realtor or financial advisor whose channel is one weekly talking take, these numbers are permission to keep it short. The sub-minute video is not the compromise format. It is the format the retention data prefers.
- 50% average engagement for videos under 1 minute, the highest of any length band, in Wistia's analysis of 14 million videos across 100,000 businesses (Wistia State of Video Report, 2025).
- 46% average engagement for 1-to-3-minute videos and 45% for 3-to-5-minute videos, from the same dataset (Wistia, 2025).
- 66% of viewers finish a business video under 1 minute long; only 22% finish one over 20 minutes (Vidyard Video in Business Benchmark Report, 2022).
- 60% of videos produced for business purposes run 2 minutes or less (Vidyard, 2022).
- Use of AI in video production more than doubled year over year among the businesses Wistia surveyed (Wistia, 2025).
Share of viewers who finish a business video, by length
Completion rates from Vidyard's Video in Business Benchmark Report, 2022.
- Under 1 minute66%
- 1 to 2 minutes56%
- 2 to 10 minutes50%
- Over 20 minutes22%
This is the whole editor
Highlight a phrase and a clip lands on those exact words. No timeline, no keyframes, no layers.
Sound off, captions on
A talking-head video is a speech, and most of the feed cannot hear it. The caption research is the least ambiguous body of data on this page.
- 69% of people watch video with the sound off in public places, in a survey of 5,616 US adults (Verizon Media and Publicis Media, 2019).
- 80% of consumers in that survey said they are more likely to watch a video to the end when captions are available (Verizon Media and Publicis Media, 2019).
- 12% longer average view time for captioned video ads, in Facebook's own internal tests (Facebook, 2016).
- 88% of TikTok users say sound is vital to the TikTok experience, the exception that proves the platform rule: caption for the feed, but do not strip the voice (TikTok Marketing Science via Kantar, 2022).
Person-led and lo-fi: the platform studies
The platforms have commissioned their own research on what style of video works in-feed. The consistent answer: native, informal, person-led, and fast to the point.
- 74% of viewers say ads made in a TikTok-native style catch their attention (TikTok Marketing Science Global Creative First Study via Ipsos, 2022).
- 3.3x more actions such as clicks, likes and shares for TikTok-first ads compared with ads ported from other platforms (TikTok Marketing Science Global Entertaining Ads Study via Marketcast, 2022).
- 90% of an ad's total recall impact is captured within the first 6 seconds (TikTok Creative Guide, 2020).
- 65% higher brand affinity and 25% higher recall when the product is shown on screen while a person talks about it (TikTok Marketing Science US SMB Creative Effectiveness Study via Lumen, 2021).
- 66% of consumers say short-form video is the most engaging type of in-feed social content (Sprout Social Index, 2022).
- 36% year-over-year growth in video watch time on LinkedIn, with short-form video creation growing at twice the rate of other post formats (LinkedIn, 2024).
The creator economy, counted
The talking head is also an industry. The people speaking to their own cameras are now measured by investment banks and platform earnings calls.
- $250 billion: Goldman Sachs' estimate of the creator economy's addressable market in 2023, projected to roughly double to $480 billion by 2027 (Goldman Sachs Research, 2023).
- 50 million global creators, growing at a 10 to 20 percent compound annual rate by the same estimate (Goldman Sachs Research, 2023).
- 303 million people worldwide identify as creators, 165 million of whom started after 2020; the US alone added 34 million (Adobe Future of Creativity study, 2022).
- 70 billion daily views on YouTube Shorts, watched by more than 2 billion signed-in users a month (YouTube, 2023).
- 200 billion daily views on Shorts by mid-2025, per YouTube's own announcement (YouTube, 2025).
- $9.9 billion in YouTube ad revenue in a single quarter, with annual revenue across ads and subscriptions passing $60 billion in 2025 (Alphabet earnings, 2026).
What the business surveys say, sample sizes included
Survey data is weaker than platform data, so the sample sizes are stated here rather than hidden. Wyzowl's annual survey polls a couple of hundred marketers and consumers; treat its figures as directional, not as laws.
- 91% of businesses say they use video as a marketing tool, matching the survey's all-time high (Wyzowl State of Video Marketing, 2026).
- 85% of people say watching a video has convinced them to buy a product or service (Wyzowl, 2026).
- 95% of video marketers call video an important part of their strategy, in a survey of 205 marketing professionals and consumers (Wyzowl, 2025).
The famous statistics we refuse to publish
Three numbers dominate every other page about video effectiveness. We traced each one to its origin, and none of them survived. They are described here so you recognise them, not so you repeat them.
Viewers retain 95% of a message in video and 10% in text. There is no peer-reviewed study behind this. Insivia, the agency the figure is usually credited to, has said it came from its own small, directional survey of roughly 200 B2B buyers in the late 2000s. It is also sometimes pinned to Edgar Dale's Cone of Experience, which never contained percentages at all.
One minute of video is worth 1.8 million words. This is arithmetic, not measurement: Forrester's James McQuivey took the adage that a picture is worth a thousand words, multiplied by 30 frames a second, then by 60 seconds (Forrester, 2008). McQuivey meant it as a metaphor. It is cited as if someone counted.
The average attention span is 8 seconds, less than a goldfish. This traces to a 2015 Microsoft Canada consumer report, which took the number from a site called Statistic Brain. When the BBC went looking in 2017, Statistic Brain could not produce a source, and attention researchers said there is no evidence human attention spans have shrunk at all. The goldfish part is wrong too: goldfish are used in memory research precisely because they remember.
The sources
Primary sources for every number above, so you can check them the way we did.
- Crouzet, Kirchner and Thorpe, Fast saccades toward faces, Journal of Vision, 2010.
- Bakhshi, Shamma and Gilbert, Faces engage us, CHI conference, Georgia Tech and Yahoo Labs, 2014.
- Guo, Kim and Rubin, How video production affects student engagement, ACM Learning at Scale, 2014.
- Garcia and Yousef, study of talking heads and annotations in video lectures, 2023.
- Wistia, State of Video Report, 2025. Vidyard, Video in Business Benchmark Report, 2022.
- Verizon Media and Publicis Media, caption survey of 5,616 US adults, 2019. Facebook caption tests, 2016.
- TikTok Marketing Science studies: Ipsos 2022, Marketcast 2022, Lumen 2021, Kantar 2022, and the TikTok Creative Guide, 2020.
- Sprout Social Index, 2022. LinkedIn video announcements, 2024.
- Goldman Sachs Research, 2023. Adobe Future of Creativity, 2022. YouTube announcements 2023 and 2025, Alphabet earnings, 2026.
- Wyzowl, State of Video Marketing, 2025 and 2026.
Where Cutroom fits, priced
The numbers above describe one practice: a real person, on camera, under a minute, captioned, posted weekly. Cutroom exists to make that practice survivable after the filming.
Film one talking take on your phone, up to 3 minutes. Cutroom transcribes every word, writes the hook, places b-roll on the exact words that describe something, and burns in captions for the 69% watching on mute. The finished 9:16 MP4 is back in about a minute, and you direct on the transcript, never on a timeline.
A batch is 100 credits and export adds 20 credits per finished minute, so a 30 second video comes to about 110 credits. The first 3 finished videos are free in your first 7 days, with 600 credits and no card, and export with a Cutroom mark across the middle. Lite is $19.99 a month for 1,250 credits, about 11 finished 30 second videos, with nothing on the picture. Cancelling is one click under Settings.
What it will not do: sources over 3 minutes, landscape exports, or a promised result. The research says short, captioned and human wins on average. No tool can promise your average.
Questions people ask
- Are talking head videos more effective than faceless content?
- The strongest evidence says faces attract and hold attention: photos with faces earned 38% more likes across 1.1 million posts (Georgia Tech and Yahoo Labs, 2014), the eye reaches a face in about 100 milliseconds (Journal of Vision, 2010), and informal talking-head lectures out-engaged studio versions across 6.9 million sessions (ACM Learning at Scale, 2014). None of that guarantees any single video wins.
- How long should a talking head video be?
- Under a minute for social. Videos under 1 minute average 50% engagement, the highest band Wistia measured across 14 million videos (2025), and 66% of viewers finish a business video that short (Vidyard, 2022). Engagement falls off sharply past about 6 minutes (Guo, Kim and Rubin, 2014).
- Is the statistic that viewers retain 95% of video messages real?
- No study supports it. Insivia, the agency it is credited to, has described it as the pattern from its own small survey of about 200 B2B buyers, not peer-reviewed research. This page does not publish it, and pages that do are copying each other, not a source.
- Do people watch talking head videos with the sound on?
- Often not: 69% watch with sound off in public places, and 80% say captions make them likelier to finish a video (Verizon Media and Publicis Media, 2019). TikTok is the exception, where 88% of users call sound vital (Kantar, 2022). Caption everything and keep the voice.
- Where do the statistics on this page come from?
- Peer-reviewed studies, platform-published data and named surveys, each cited inline with its year. Where a survey is small, the sample size is stated. Recycled figures with no traceable origin are listed in their own section and excluded from the counts.
- What does a weekly talking head video cost with Cutroom?
- About 110 credits per finished 30 second video: a 100 credit batch plus an export at 20 credits per finished minute. The first 3 are free in your first 7 days, with 600 credits and no card. Lite is $19.99 a month for 1,250 credits, about 11 of those videos, and cancelling is one click. Sources cap at 3 minutes and exports are 9:16 MP4.
Every number above has a name and a year on it. What they point at is one practice: a face, under a minute, captioned, every week.