The live cam presence dataset

What is in it

One row per scope per hour of the day. A scope is the whole site, one network, one country, one gender or one category. For each we publish the average number of performers online in that hour, the number of performer-hours that average was measured from, and how many days the window covers.

The hour is UTC. avg_online is the average count of distinct performers seen online in that hour across the window, rounded to a whole performer. observations is the sample behind it: the total performer-hours recorded. sample_days is the number of days in the window.

A figure without its sample size is not evidence, which is why observations is in the file rather than only the average. Scopes measured fewer than 30 times across the whole day are left out entirely: the shape of a day measured a handful of times is noise, and publishing noise as data is worse than publishing it as a picture.

What is not in it

Nothing that identifies a performer. There are no usernames, no room identifiers and no per-person rows. Every figure is a count of people, never a record of one. We hold presence per performer internally to draw the ribbon on their own page, and none of it is published here.

Location and gender are as performers set them on their own network profile. Neither is verified by us, and location in particular should be read as a claim rather than a fact. Two of the networks we list publish no location at all, so they contribute to the network and gender scopes but not the country ones.

How it is collected

Each network publishes a feed of who is streaming right now. We poll those feeds on a schedule, record who was online in which hour as a 24-bit map per performer per day, and roll the maps up nightly into the counts above. Nothing is scraped, nothing is inferred, and no row is written per poll. Performers under 18, or whose age a network leaves unstated while flagging it, are rejected before anything is stored.

The counts are of performers we list, which is the public, free-to-watch part of each network. A network's own total will be higher, because it includes rooms that are private, offline or excluded from its feed.

Known limits

• A performer streaming on two networks is counted on both. We do not attempt to match identities across networks and would not publish it if we could.
• Feeds go down. An hour when a network's feed failed reads as a quiet hour, not as a gap.
• The window is rolling, so re-downloading the file later gives different numbers for the same hour. The computed_at column dates each row.
• Country coverage is partial and uneven. Roughly one performer in eight declares something that is not a country at all, and those are excluded rather than guessed at.

Licence and citation

The dataset is published under the Creative Commons Attribution 4.0 International licence. Use it for anything, including commercially, as long as you credit the source. A citation we would recognise:

CherryLiveCams, Live cam presence dataset, 2026. https://www.cherrylivecams.com/data

If you are writing something that uses it and want a figure checked, a scope that is not published, or a longer window, ask us. We would rather the number in your piece were right.