TagCow LogoTechlife had the chance to try out TagCow, an automated photo tagging service, while it was still in beta. TagCow has now emerged from beta and is open to anyone. Automated photo tagging is equivalent to Google Search for your personal digital photos. From professionals to amateurs having quick and easy ways to sift through millions of images that are accurately tagged makes this service truly amazing.

What makes TagCow a bit more unique is the method of tagging, it is automated. Other services such as Picasa or Flickr rely on your efforts to tag photos, while Google is using a pseudo-game to help improve its tagging search results. This is what sets TagCow apart.

Tagcow’s “how” tagging is accomplished is a bit hidden on the website, but in speaking with Michael Droz, CEO and co-founder, he revealed it is part automated and part human-based. I would even venture to suggest different humans were assigned to my photos, as the results were slightly different for each image I uploaded. There was even a typo! Due to this tagging says it will take 24-48 hours, mine took about 55 hours.

TagCow's automated tags

TagCow has some really unique features such as people tagging. You can tell the service who a specific person is with a base photo and then all future photos will be tagged with the same person. Same goes for descriptive tagging of other source objects, locations, etc. This makes the process for the user very easy. It is certainly a very smart way to approach photo tagging.

TagCow's automated tags with a bit of an error.

My beta of the service really put it to the test. I used a lower resolution camera from my Sidekick 3 taking photos from a visit to the traveling Star Wars museum exhbit.


  • Tagging and then searching is very useful for managing your digital photo collection.
  • Automating the process even with small mistakes is a huge time saver.
  • TagCow hit a homerun with the depth of the knowledge of their human taggers, with accurate labels of different ships from the Star Wars saga, including a Trade Federation Tank, a Podracer, a Landspeeder, an X-Wing Fighter, the Millennium Falcon and an Imperial Destroyer. It’s only mistakes were calling an AT-AT a “fighter” and a mistype of “X-wing” as “X-wind” in one tag.
  • TagCow’s tags are portable and actually use standard technology to tie them to the photo.


  • Inability to edit, add, or remove tags on the TagCow site is a big negative.
  • The bulk uploader is great, but there should be a bulk download feature as well.
  • These two are easily fixable, the last con is pricing. Coming out of beta they decided this service was worth something, and likely their human taggers agree. Their price is per photo, starting at $9.95 for 250 up to 2500 photos for $89.95. They do offer a free 100 photos to get you started, but this pricing model is a big problem, as it starts to make the average consumer question how important are automated tags.
  • Lack of an ad supported model, this would help offset costs, but as popularity increases having humans involved will always make scaling the business more challenging.
  • Accuracy is still questionable. One photo of Yoda was tagged “alien, star trek” while the next photo was accurately tagged “yoda, star wars, toy”. Improving this will improve the service. Another photo had a boy with a Stormtroper and was labeled “boy, storm trooper” the next photo labeled, “young boy, star wars character”. If the same human did both photos they would have both been tagged more closely.

TagCow results in Picasa
Overall: I would have liked to see TagCow stay in beta and add a few more bells and whistles before going live. Should they implement user tag correction, my suggestion would be to track changes or corrections per tag, tag deletions and tag additions to give users and themselves an accuracy report card. By quick glance of my photos they approached the 90%+ accuracy rate, yet as humans we focus on the 10%. Finally, photo tagging by automation and humans is a great leap forward for the industry, but I question if the pricing model can support the growth needed. Does it scale?