One fingerprint of unmodified GAN-generated faces (at least, those made with widely-available tools) is that the major facial features (especially eyes) are in the same location on every image. This becomes obvious when one blends multiple GAN-generated faces together.
There are other signs that @prof_freedom's profile image is not a real photograph of a person's face. The most obvious is the unrealistic teeth; there are also some strange things going on where the hair meets the background.
The use of a GAN-generated face isn't the only odd thing about @prof_freedom (permanent ID 1252542024299012097). The account has been renamed at least once, from @/DaFeid (archived here: web.archive.org/web/2020121719…). Also, at least some of @prof_freedom's photos are stock photos.
Misleading tweets about masks, vaccines, and other COVID-related topics from @prof_freedom have frequently gone viral.
(Note: while this thread focuses on English content, this account also tweets frequently in German. Topics appear to be similar to the English tweets.)
One of @prof_freedom's least endearing habits: spamming people's replies with the same images over and over. @prof_freedom has replied hundreds of times to a variety of users with variations on the same nonsensical "flowchart".
Additional tips on identifying GAN-generated faces:
The GAN-tastic HR professionals are part of a network of 53 accounts created between May and July 2022. All have GAN-generated face pics, and none has tweeted since August 12th. When active, they tweeted exclusively via the Twitter Web App and took at least one day a week off.
Many of the images used by this network have been stretched or cropped, preventing them from being flagged as GAN-generated by facial feature position. They still contain anomalies, however: disappearing glasses, mismatched earrings, and surreal clothing/jewelry/backgrounds, etc.
This fake follower botnet consists of 23965 accounts created between August 3rd and August 7th, 2022. None of these accounts has ever tweeted or liked a tweet. All use the same naming scheme and most have repetitive biographies.
In addition to the aforementioned repetitive biographies, this fake follower botnet also uses the same profile pics on multiple accounts.
It's unclear why anyone would want to spend money a bunch of used cryptocurrency/NFT Twitter accounts, but a user with the descriptive name of "VIP STORE" is selling 21 of them on shady account sales site accs-market(dot)com. #ExtremelyUnwisePurchases
The 21 accounts being sold by "VIP STORE" are all older accounts, created between 2008 and 2014. At least 13 (and possibly all) have been renamed, with the new names all containing "nft", "btc", or "crypto". (The old names were found in data we downloaded for previous projects.)
Although these 21 for-sale accounts have existed for years, they appear to have gained most or all of their followers in 2022. (We know this because each account has early followers created in 2022, which means all subsequent followers followed them in 2022.)
One telltale trait of unmodified GAN-generated faces (at least, those in common use) is that the primary facial features (especially eyes) are in the same position on every image, regardless of the angle of the head. This becomes obvious when multiple images are blended together.
There are several additional signs that @youpress19801's profile image is GAN-generated, including the surreal background and the random fragments of clothing and hair hanging in midair to the right of the face.
First, we have @GypsyCeltic (ID 1307840989617434624), an account with a GAN-generated face and a propensity for posting false statements about COVID vaccines. This allegedly Florida-based account is followed by Ron DeSantis campaign operative @ChristinaPushaw. #FloridaGAN
Next up is @dogoym (ID 1346315171229143044), an antisemitic troll account with a GAN-generated pic and a name that references the antisemitic dog whistle "The Goyim Know". According to one of this fake account's replies to Elon Musk, ~30% of Twitter accounts are fake.
It's a Friday in August, and a bunch of spammy accounts are tweeting links to what appears to be the same article about recreational marijuana and auto insurance rates in Missouri on a variety of obscure websites.
Although the article has the same title on each website, the text varies slightly from site to site (possibly a result of article spinning: en.wikipedia.org/wiki/Article_s…). The original version appears to be from St. Louis news outlet @KMOV.
The articles are being shared on Twitter by a network consisting of 13 automated Twitter accounts, each of which tweets links to a different website. Most of the accounts (10 of 13) were created within the last month, and all but one have the biography "all news for you".