The Internet Archive, in partnership with George Blood LP, has been working since 2016 to digitize thousands of 78rpm records, some of which were made in the early 1900s. Over 250,000 records have been preserved, and the archive posted a Twitter thread yesterday detailing how each record is cleaned and filed away.

Each record is cleaned on a machine that sprays distilled water onto its surface. A little vacuum arm then sucks up the water, along with whatever dirt and nastiness has built up in the record’s grooves over the years.

The discs are then photographed, and the photos are referenced to pull info from the discs’ labels and add it to the archive’s database by hand. While most of the records are from the biggest record labels like Columbia, RCA Victor, and Capitol, the archive has found 1,700 other labels.

Some of the 78s are a treat to look at, including picture discs that have an image printed across the entire surface of the record. I recommend “Vem Vem (The Cuban Kissing Game)” from 1947 and “Night Herding Song” from 1949.

78s don’t have standardized groove sizes, so recordings taken with various stylus tips will each sound slightly different. Here, you get four: the records are digitized with custom-built turntables that have four arms, each with a different stylus. Each version is available to download separately, with the file name showing “3.3_CT” for a recording from a 3.3 mil truncated conical stylus and so on.

Four styli might make you nervous when we’re talking about preserving fragile records, some of which are over a hundred years old. But most of these 78s are made of hard shellac, which makes them prone to shattering if mishandled, but is also tough enough that the grooves won’t be damaged by a stylus. Modern equipment with diamond tip styli also exert much less force than old players with steel needles, George Blood LP points out.

The whole collection is a little overwhelming to dig through, but you can narrow your browsing by year, genre, and artist. If you’d prefer to be surprised, the @great78project Twitter account posts a new record from the collection every hour.