Real numbers on every YouTube page.
Lens shows what a video actually did against its own channel, what a channel’s health is, and why a title worked — on YouTube itself, while you browse. No credit wall, no guessed numbers, nothing to set up.
Add Lens to Chrome — freeWhat Lens is
A measuring instrument for YouTube, built by a growth agency for its own research and given away.
Most tools show you views. Views tell you almost nothing: 50,000 views is a disaster for one channel and a career-best for another. Lens shows every video’s multiple — how it did against what its own channel normally does — so a breakout looks like a breakout whatever the channel’s size, and a flop on a huge channel looks like the flop it is. Around that one number Lens adds the channel’s health, the video’s speed, and the title decoded against everything measured in its category.
The one rule
The scale
The multiple compares a video’s views with that channel’s own normal — Shorts against Shorts, long-form against long-form, so a long video is never judged against Shorts. (2.0.43) A video counts as a Short only if YouTube lists it as one. (2.0.44)
| Badge | Multiple | Meaning |
|---|---|---|
| ≥5× | 5.0 and up | Breakout. The packaging or subject reached far beyond the channel’s usual audience. Worth studying closely. |
| ≥2× | 2.0 – 4.9 | Winner. Twice the channel’s norm or better. |
| 1–2× | 1.0 – 1.9 | Par. Did what this channel usually does. |
| <1× | under 1.0 | Under. Below the channel’s own norm. |
| New | under 2 days | Too new to grade. Early views are arithmetic, not a signal. |
| 🕒 | 2 to 14 days | Moving now. A grey, provisional pace against the same channel’s uploads at the same age. It sits beside the multiple and never replaces it. |
On feeds, search results and channels
- Why it was built
- A viewer meets your video as one tile among many, at feed size, against whatever else is on screen. A draft that looks fine at full size on your desktop can vanish there. Preview-only tools show you the swap; none of them says anything about it.
- How to use it
- Search the topic your video competes in, or open the home page, and press the pill. Click to add a PNG, JPG or WebP draft, type the title, press Preview in feed, then scroll and squint. Reset puts the real feed back, and so does leaving the page. With a VideoCore account, press AI score this packaging.
- What it gives you
- The planted feed, and with an account a score out of 100, a predicted click-through band, one blunt verdict, up to three strengths and three concrete fixes, then the repeat check with the words it read off the thumbnail. Ten AI reads a day; the same draft and title read again within a week comes back from the cache and costs nothing.
- Measured
- The score and the band are an AI estimate of click-worthiness and are labelled as such; YouTube exposes no click-through rate to anyone outside the owner’s Studio, so nothing here is a measured CTR. The repeat check is counted: the words on the thumbnail that already appear in the title, small words ignored. The image is shrunk to 640×360 in your browser before it is sent, and nothing leaves the page until you press the score button.
- Measured
- From the channel’s last 25 uploads and its tracked history. Similar channels come from measured subjects and fields, not a keyword search. The strip carries no earnings figure (Channel Compare keeps one, marked as a projection) and no openings record.
🎤 Openings on a channel page (2.0.14) is the channel’s own record by the kind of opening its videos use, read from their transcripts: the share of settled uploads that beat 2× its normal with that opening, against the share without it. It is measured on that channel only and says nothing about any other. YouTube is adding a Studio test of up to three cuts of one video; this is the measured ground such a test stands on.
On a video
A normal video, a Short, or a live stream or premiere opened at youtube.com/live/… (scored since 2.0.47; before that the live address was not recognised as a video page).
No captions yet (2.0.48). When YouTube’s own CC button says captions are unavailable, the transcript tool and the AI summary say so within a second instead of waiting for a transcript panel that will never appear. Long uploads can take days to get automatic captions; try again later. On uploads with several audio tracks (dubbed videos) the player can say unavailable while captions exist, so from 2.0.50 that verdict is used only when the page also shows no transcript button or caption list; otherwise the transcript is read as normal.
- Why
- It answers the question every creator asks after a hit: was it the subject, or the way it was said? The decode shows which parts have a track record in this category and which were a one-off.
- Why it was built
- A big multiple says a video won; it does not say what made people click. This is the first guess at that, so you can ask the same question of your own packaging.
- How to use it
- Read it next to the thumbnail: does the picture really hold something back? If you borrow the idea, borrow the reason, never the picture.
- Measured
- Nothing. It is our reading of the picture and the title, made once per video, and the panel says so: “our reading of the picture and title, not a measurement”. It is not a score and does not predict views.
- How to turn it off
- Open the Lens popup, Outlier frames and badges, and untick Colour frame around the thumbnail. Tick it again to bring it back; Lens remembers your choice.
- Measured
- Nothing new: the outline is a colour for the same multiple the badge shows.
- Why it was built
- To see what a channel’s winners do differently, and what a video has tried, without leaving YouTube.
- Measured
- Openings, thumbnail versions, links and comments are counted or listed, not scored. Hooks, Thumbnail read and Field read are a model’s reading of those counts and pictures; they say so, carry their counts, and say plainly when nothing separates the winners from the rest. Not a prediction, and a channel’s own history, not a rule.
- Measured
- The field is the channel’s own: the competitors it is swept with, or else the channels measured as nearest to it (they share its subject phrases, sit within ten times its size, and have ten or more measured uploads), named on the line as “X and 14 channels near it”. Never a YouTube category. A channel with too few neighbours in the database gets structure only, from the whole database.
- Why it was built
- The plan you made in VideoCore should be in front of you while you fill in Studio, without a floating box over the form.
- Measured
- Nothing here is scored. The panel reads only the tab’s address (the video and channel ids) and asks VideoCore for your own plan; it never reads Studio’s page. Sign in with VideoCore to see it.
New in 2.0.34
Three additions. Each one says why it exists, how to use it and what you get from it.
- Why it was built
- To see which part of a video is the part people come back for, and what else the channel did around a jump.
- Measured
- The curve is YouTube’s own “Most replayed” graph read from the page, a shape not a count: the percentage is against the graph’s highest point, usually the opening, which is never counted as a peak. Peaks are picked by VideoCore. Nothing is stored. A video with no graph shows nothing.
- Why it was built
- A video that is re-titled and re-thumbnailed a few times in its first days is hard to learn from: the numbers belong to several different packages.
- Measured
- Counted from the changes Lens detected. It is the date Lens saw a change, which can be a little after it was made, and it shows only when Lens first read the video within 12 hours of the upload; otherwise it says nothing rather than undercount. No line means no change was seen.
- Why it was built
- A video needs 14 days before its multiple means anything, so the uploads from the last two weeks had nothing to say. Other tools fill that gap with “views per hour”, which mostly tells you a big channel is big. This tells you a video is fast for its own channel.
- How to use it
- Search a topic or open a channel, then open the ▶ Tools tile. The row lists up to four videos, fastest first. Hover a row for exactly how it was measured. Click it to open the video.
- What it gives you
- What is moving this week in a niche, before the outlier badges can tell you. It is labelled provisional and stays grey on purpose: it ranks what is moving, it does not predict a winner.
- Measured
- The video’s views so far against the same channel’s own uploads at that same age. Where a channel has too little of that history yet, Lens shows nothing on the card and the panel says “not measured yet”. It does not fall back to views a day against older uploads, because a young video always looks fast against older ones and the number would mislead. Tracking started on 21 September 2026 and fills in every night, so coverage grows on its own. The Tools row lists only videos running ahead of their channel. How often a fast start ends a winner is not measured yet; that test starts on 6 October 2026 and the panel says so until it has run.
- Why it was built
- People look at a thumbnail in Lens and want to make a better one; the Designer is where that happens, one click away instead of a search.
- Measured
- Nothing. It is a plain link: Lens passes only the title, never this video’s picture or another channel’s look, and nothing of the Designer is copied into the extension.
- Why it was built
- A multiple against older uploads can move for counting reasons alone; the note says so instead of letting it pass as a win.
- Measured
- Nothing new. Only the wording changed; engaged views, which earnings use, stay visible only in a channel’s own Studio.
- Why it was built
- A thumbnail test has losers, and the losers teach as much as the winner. Until now the panel showed the last five swaps and the archive read a video at most twice a day, so a three-way test could be recorded as two.
- Measured
- Read from the public thumbnail file and stored by content, never redrawn. A Studio test hands each viewer one of up to three thumbnails from its own signed address, which no server can read, so a variant reaches the archive only when a Lens user’s feed shows it: the card reports the public picture address, nothing about the viewer, and the version appears as 🧪 test #N. Ordinary swaps are also re-read every ten minutes. A version nobody with Lens was shown was never seen; the line says so rather than claiming a complete record.
- Why it was built
- YouTube has announced thumbnails that change with the viewer, up to three on one video, and Studio already tests three today. A face and a running joke pull in a fan and mean nothing to a stranger, so one thumbnail cannot do both jobs.
- How to use it
- In VideoCore open Thumbnails, pick your project and press Write three thumbnails. Build each one on the canvas. Then choose how to test: run them in YouTube Studio under Test & compare and type in the watch-time shares when it ends, or let VideoCore run the test for you on a connected channel. Lens then shows them on that video, on YouTube and inside Studio. You can remove any one of the three and still test the other two.
- What it gives you
- The three options side by side on the video itself, each with its measured share and a plain sentence on which one led and by how many points. A lead under three points is called too close to call.
- Measured
- Two ways. Studio’s shares of watch time, typed in by a person, because no API exposes a thumbnail test. Or VideoCore’s own test: each thumbnail is live for a few days in turn and the video’s real daily views are counted; it only calls a winner on a lead of 10% or more, because the days differ. Lens estimates nothing here. Only the channel’s own VideoCore account is shown its tests. A title test on the same video waits until the thumbnails have finished, so the two results stay readable.
- Why it was built
- A viewer gets two looks at your video in a feed: the picture and the title. If both say the same words, the second look is wasted. It is one of the most common faults on small channels and it is easy to count.
- How to use it
- Upload a draft thumbnail with its title in the Thumbnail test on any feed or search page. The card reads the words on the image and compares them with the title. Nothing else to do.
- What it gives you
- A clear yes or no, with the shared words listed, and what to put there instead: the result, the number, or what is at stake.
- Measured
- Counted, not judged: how much of the thumbnail’s wording already appears in the title, ignoring small words. A one-word thumbnail is never flagged.
Since 2.0.34
What changed on the page you use every day. Each line says what you see and what to do with it.
- Why it was built
- Saving or sharing the video you are watching used to mean going back to a thumbnail grid.
- How to use it
- Hover the player, press 📸 for a shareable image of the card or 💾 to file the video with its numbers.
- Why it was built
- A button that does nothing visible looks broken whether it worked or not.
- How to use it
- Watch the icon after you press. A tick means done; a warning means try again, or check you are signed in for the save.
- Why it was built
- The notice was larger than it needed to be, and the repeated asks used the preview up for nothing.
- Measured
- From the channel’s own history of views at each age. The band is the typical range, not a forecast.
The opening check
How a video opens, checked against its own title. It is in the Lens panel.
- Why it was built
- Someone who clicks on a title expects to hear it straight away. An opening that starts somewhere else leaves them wondering if they are in the right video. Advice everywhere says “say your title in the first 30 seconds”; this counts whether a video does.
- How to use it
- Open the Lens panel on a video. If the line shows a warning, the title’s subject is mostly missing from the opening. On your own video it says which words to say. In VideoCore the same check runs on your script in Video Editor and in Packaging Coherence, before you record.
- What it gives you
- A plain count, for example “The opening says 3 of the title’s 5 key words. Not said: cost, own.”
- Measured
- Counted from the words in the first 30 seconds of the transcript. Filler words are ignored and a number counts whether written as a digit or a word. No transcript held, no line. It is a count, not a score, and it does not predict views.
The Tools tile
One ▶ Tools launcher bottom-right on every video. Seven tools, each with a hover explanation and grouped by what a click does: into your VideoCore (red), a download to your computer, shown on this page (blue). Switch any of them off in the popup.
Post the facts
New in 2.0.55. A video’s substance as timestamped lines, checked against the transcript, ready for the comment box. You post it.
- Why it was built
- Most videos hold their facts back and make you sit through the whole thing. A comment that lists them with their moments gives everyone who reads the comments the substance in a minute, gives the video clickable chapters, and says where the facts came from.
- The text
- The gist, one
m:ss factline per fact (YouTube turns each timestamp into a link), and nothing else: no hashtags, no emoji, no asks. The credit line is off by default (2.0.106). If you turn on Add a credit line to Post the facts comments in the toolbar popup, a blank line and a last line are added, always exactly: Provided by the VideoCore Lens Chrome extension: videocore.uk/lens. It is off for new installs and for every install that has not turned it on; the caps, the kill switch and one comment per video are unchanged. A video with fewer than three checked facts gets no comment text at all. Copy the text puts it on your clipboard. - Put it in the comment box
- When this button is on, it writes exactly that text into the page’s own comment box, the top-level one, never a reply, and stops. Read it, change or cut anything, then press Comment yourself. It is offered once per video, five times a day per install, and the server can switch it off for everyone at once without an update; for the first week after release it is off while we watch how the first posted comments fare, and the panel says so. If YouTube shows no box (signed out, or comments off) the panel says so and you can still copy the text.
- What it never does
- Post. Lens never presses Comment, never fills a reply box, never calls YouTube’s comment API and asks for no YouTube permission; a check in every build fails if that ever changes. Only you can post, and you can edit or delete the text first. The closing line is never reworded or hidden.
- Measured
- Read from the video’s transcript by one model call on the server. Then checked in code: a fact whose time is not in the video, or which states a number the transcript does not say, is left out, and the panel says how many were. A video read by anyone in the last 30 days is served from the cache, instant and free. Each install may read ten new videos a day, and the panel says where you stand; there is a daily ceiling for everyone as well. Over an hour, only the first hour is read, and the panel says so. A video without a transcript says so.
- Privacy
- The read and the box-fill ask send a random install id (made once, not tied to you or an account) so the daily counts and the once-per-video rule can be kept per install. Nothing else carries it; clearing the extension’s data makes a new one. Which videos were filled is kept so posted comments can be checked for a month; no comment text or account is stored.
The popup
The ▶ icon in Chrome’s toolbar. It reads the page you are on.
This page shows the numbers already on screen: on a video, its multiple, views, per hour, engagement, age and comments; on a channel, its health, subscribers a day, uploads a week, best recent outlier, biggest lever and similar channels; on any list, how many videos are rated and how many sit in each band. Tools on this page lists every tool that applies here as a button. Settings fold below: the badges, the comment counts, and each tool on or off.
Five ways to use it
Install and settings
Card icons (2.0.60). In the popup's ⚙ What shows on YouTube card (visible, not folded, since 2.0.63) one master switch turns every thumbnail icon off or on together (2.0.61); below it you can switch off the 📸 download-card icon and the 🤝 collab chip that sit at the top-left of thumbnails, each on its own. Off takes effect on the page at once; the remaining icon moves into the first slot.
Questions
Where do the numbers come from?
Public YouTube data, measured against each channel’s own normal — using VideoCore’s proven formulas, the same ones used everywhere, so the extension and the platform never disagree.
Why does a collaboration video score so high?
YouTube counts a collaboration (“X and 2 more”) as the host channel’s upload and shows it to every collaborator’s subscribers, while the multiple compares it to the host channel’s own uploads, which never had that reach. The 🤝 card under the headline names every channel on the video and their subscribers together, and the figures that are host-only are labelled so. Read the multiple as “against the host’s usual”, not as the packaging alone.
Do I need an account?
No, but a free one shows you the numbers. Without an account, Lens (version 2.0.109 or later) tells you where each video sits against that channel’s normal in five words: well above, above, about, below or well below typical, a free preview of 150 videos and 10 channels opened in the panel and channel view for each install (the scores on the thumbnails keep showing), and Post the facts reads ten new videos a day per install. Signing in with a free VideoCore account shows the exact multiple, the share of typical, the rank, the typical range and the history, adds the AI tools (thumbnail scoring, clipping to your Swipe File) and removes the daily limit. New accounts are not open yet; until they are, Lens points you to the free channel check instead, which needs no account. Older versions of Lens still show the exact numbers until they update.
Why is there no click-through rate?
YouTube does not expose impressions or CTR for any video to anyone outside its owner’s Studio. A CTR shown by any extension is an estimate. Lens shows what is measured and says what is not.
Why does a new video say “New”?
Because a two-day-old video’s multiple is meaningless: the video has had no time yet to be compared with its channel’s normal. Lens shows the tracking rate and waits.
What does the 🕒 chip mean?
The video is between 2 and 14 days old, and this is how fast it is going against the same channel’s own uploads at the same age. It is provisional and grey because a video that young has not settled. Use it to see what is moving this week, not to call a winner.
Is the Thumbnail test score a real click-through rate?
No. It is an AI read of your draft and title as a pair, at feed size, and it says so. YouTube gives no click-through rate to anyone outside the owner’s Studio. The one counted line on that card is the repeat check: how many words on the thumbnail already sit in the title. Use the planted feed with your own eyes first; the score is a second opinion.
Can Lens show which thumbnail is winning my test?
It shows the result once you have typed it in. YouTube does not let any tool read a thumbnail test, so Lens cannot fetch it. When your test in Studio finishes, type the watch-time shares into VideoCore (Tests) and Lens shows them on the video with the difference between them.
Is it really free?
Yes, with no credit wall and no meter. VideoCore’s paid plans exist for creators who want the full platform; Lens does not nag while you research.
Does Lens keep working in a tab I left open?
No. From version 2.0.162, while a YouTube tab is behind another tab, minimised or on a locked screen, Lens scores nothing and asks for nothing. When you come back to the tab it catches up once, so what you see is up to date.
What does Lens collect?
It reads public YouTube page data to compute its scores. Your browsing is not tracked or sold. The on-device summary never uploads the video.
Who makes it?
ViewsToClients, a YouTube growth agency headed by Jake Trinder, and VideoCore, its platform. Lens is the instrument the agency uses for its own research.
Measures
Every figure in Lens is worked out with VideoCore’s own proven formulas from public YouTube numbers. They are VideoCore’s calculations, not YouTube metrics.
| Term | What it tells you |
|---|---|
| Multiple | How the video’s views compare with that channel’s own normal, Shorts and long-form kept apart. |
| Views per hour | Views ÷ hours since publish. Measured, never averaged from a model. |
| Engagement | (likes + comments) ÷ views, as a percentage. |
| Health | 0–100, a summary of how healthy the channel’s recent numbers look, from measured figures only. |
| Winner share | Of the videos in a category wearing a pattern, the share that beat 2× their own channel. |
| Lift | A pattern’s winner share ÷ the field’s overall winner share. 1.9× means titles with that construction win almost twice as often. |
| Moving now (🕒) | For a video 2 to 14 days old: how its views so far compare with the same channel’s own uploads at that age. No fallback: a channel without enough of that history tracked reads “not measured yet”. Always provisional. |
| Share of watch time | For a thumbnail test: the share YouTube Studio reports for each thumbnail, typed in by a person. The lead is the top share minus the second. Under three points is too close to call. |
| Repeats the title | How much of the thumbnail’s wording already appears in the title, small words ignored. |
| Similar channels | Channels in the same measured field and of a similar size, or whose videos cover the same measured subjects. |
| Named in | Other channels’ videos in the last year whose title names this channel: collabs, guest spots, interviews, stories. The one shown first is the one furthest above its own channel’s normal. Counted from the database, so a channel nobody has measured shows nothing. |