Atmosfy's Atlas repackages its consumer video map of 100 million places as licensed AI data

Atmosfy's own totals imply an average of five experiences per place: more than 500 million spread across 100 million places ^[15]. The same map claims 20,000 cities and 150 countries ^[3]. An average over that much territory says little about any one venue. It cannot show whether clips cluster in a few dense neighborhoods or spread thinly at one per listing.

Freshness depends on that same distribution. Atmosfy says its Spectra model treats each video as a time capsule tying a place to a moment, and that together the clips keep its view of places up to date ^[5]. A time capsule is dated by design: Spectra can tie a clip to a place and a moment, but the picture of that place only changes when someone films it again. Atmosfy has been gathering this kind of footage since at least the pandemic, when its pitch was that customers could post videos showing local businesses were open and worth visiting ^[13].

Place matching is a separate join, and it has its own error rate. Atlas materials call the corpus geo-tagged and structured ^[6]. A geo-tag is a coordinate, and resolving it to the right venue in a building with three tenants takes an entity-matching step. In a recommendation product, a wrong match produces a wrong answer. For the corpus to carry over into a buyer's own place graph, I'd expect each record to need an upload timestamp, a stable place identifier and a match confidence the buyer can filter on.

The rights layer is the best-documented piece. The sublicensing grant in Atmosfy's terms of service names AI development and model training explicitly ^[10]. The drafting is specific, and I think it is the strongest part of the package. The announcement did not disclose how many videos are recent or license-ready, and the terms leave open how much of the claimed corpus they cover and what any customer license allows ^[4][12].

Founder and CEO Michael Ebel built Atmosfy for people choosing where to go. Atlas sells to companies building systems that make those choices for them ^[2]. A Berkeley Haas profile says he came to video while at Meta, seeing it as a fuller way than static ratings and text reviews to show what a venue was like ^[14]. Atmosfy's website shows Atlas in a travel-discovery setting ^[9]. The company's seed round has two public sizes: $12 million in Atmosfy's August 2023 announcement, led by Redpoint Ventures, and $14 million in a Haas magazine account from the same time ^[16][17]. The two figures differ by $2 million ^[19]. New backers named at launch include Capital One Ventures, DNX Ventures and Suntory Ventures ^[18].