Twelve Labs Word-Based Video Search Raises $17M
Started by a group of AI and software engineers who met at Korean Cyber Command, Twelve Labs offers a platform for word-based video search.
In this report, based on an interview with CEO Jae Lee, Head of Business Soyoung Lee, and Head of U.S. Operations Anthony Giuliani, IPVM examines the company's background, positioning, approach, and outlook.
**********
****** **** *** *********** ** ***** 2021 ** * ***** ** ****** engineers *** ***** *** ***** ** the **** ***** ** ****** ***** Command ***** ******** ** ***** *************, according ** *** ***. ****** ****' Head ** ******** **** **** ******** their ***** ****** *** ** ****** beta ~* ****** ***, ****** **** their "***** ** ** ***** ******** of ********, ****-********** ***** ********, **** will **** **** ****? *** **** they ** *********?"
*******
*** ******* ****** $** ******* ** seed ******* ********* *** ****** ** an *********** *******. *** ******* *** a **** ***** **** ********* ******* ***********, ***** **** ***** ** ***** 2021. *** *** ***** **** ****** the *********** *******, **** ****** **** money ** * "****** *** ******* round." ********* *** ***********, *** ************* $* ******* ** **** **************** *********** ***** ********* ** ***** ****.****** **** ***** $** ********* ******* ** *** **** ***** as * **** *********, ******* ***** funding ** **** $** *******. *** Lee ********* *** *********** *******:
** ****** **** ***** ******** **** was *** **** ***. ** **** led *** ***** *** **** *** seed *** ** **** *********, ***** round ***** ********** ************ **, ***** *** *** *****. Second *****, ** *** *******, ** Radical ******** *** *** ***** ******** up. ** **** *** ******* **** on ** *****.
**** ********** ********* ** ****** *********** *** ********* ** *** ******* Cohere.AI, *** ********* ** *********, *** CEO ** *******, *** *** ** Duo ********, *** *** *** ** Weights *** ******.
Revenue *** *********
*** ******* *** *** ******** *** revenue, *** ********* **** ****** *** closed **** **** *** **** *** months ** *** ****** *********:
** ***'* ******** ******* ** * private *******. ** **** ******** *** close **** ******* **** ****, *** so **'** **** ** ****** **** for *** ****** ***. ** *** early ********* ****** **********, ******* ** different *** *****. *** **'** **** lucky ****** ** **** *** ******* into ********** *** ******* ** **** paying *********. ** ** *** ***** a **** **** ***** *******.
*** ******* *** *** ***** * 2023 ******* **********, ****** ******* **** are ******* ** ******** *** ******* development:
** ****. *, ** **** **** idea. * ****, **** *** *******. But ******, *** ***** ** ** how ** ** *** **** *** focus ** ** ***** ******** ** tomorrow, **** ********** ***** ********, **** will **** **** ****?
*** ******* ******* * ********-**-******** ******** with *** *******, ***** ********* ***** have ** ******* *** ********* *** Twelve **** *********:
******* **** ********* **** *** ******* is ** * ***** ***** ** this ** ****** **** ****** ***** revenue, *** ** *** ******, **'** mostly ********** ** **** *** *** use *****? **** *** **** ** the ***** ******* ***** ********** **** are **** ***** ********** *** ** are ** *** **** ** **, right? ** ******* **** ******* **** early ***** *********, ** ****** ***, actually. *** ****, *** ****, **** of *** ****** ********** *********, ** takes ****, *****?
*** *******'* ******* ******** ****** ********* on ***** ** ***** ********* ******* the ***, *** *** ******* ********* pay ******** **** $*** * ***** to ******** * ****:
***'* *** ***'** * ******* ****'* building ********* ******* ** ****, *********** *******, ** **** **** **** analytics *******, *** ***'** **** ******** out, ***'** ******** **** ***** *** (minimum ****** *******), **** ***'** **** of ******** ****** *** ***** [* day], **** ********* ***** [* *****] then **** ******* **** *** ****** 100 ***** *** *****, ** * really ***** ******* ****'* ******** **** tens ** ********* ** *****, **** on * ****** *****, ** ****'** be ****, ** **** ***,*** ***** per *****. *** **** **** ****** quickly ***** ** ** * ****** $100,000 *** ****, *****. *** **** there's ****** ******* ******** **** ** can't ****** ******** **'** ******* ****, which **** ********* ***** ** ***** that **** **** ** **** **********, then ****'* * ************ ****** ********.
******* *** **** ***** *** ******** of ***** ********* *** ***** ** the ** *** ******.
*********
*** *** **** **** **** ** employees, **** * "**** ****** *********** and ******** **** ** *****." *** also ***** **** **** *** ******** out ***** **-**-****** **** ** *** U.S., *** **** **** ** ********* and **********.******** ***** ** ************* * ** ***********.
*******
****** **** ****** * ***** ****** infrastructure **** *** ***. *** ********* "extracts *********** **** ** ********, *******, sound, **** ** ******, *** ****** from ****** *** ****** ** **** specific ******* ***** ******** ********." ***** can ****** ****** ** * ********, where *** ****** **** ********* ******* the ******. ******* *** ********* ***** product:
** **** * ***** ** ** models **** **** ** *******, *********, people, *************, **** ** ******, *** so **. *** *** *** ***** that ******* **** ** ******** ******* relationships ******* ***** ******** ****** **** video, *** **'* *** ************ **** gives ******* ** *** ******* ** what's *********. *** ** ********* **** context **** *******, ***** *** ******* that *** ******* ******** ***, *** then ***** *** ****** ****, ***** is ** ****** ** **** *** search ***. ** **** **** ***** or **** **** ** ****** ******* category **** ******* ****** *** ** mash **** **** *** ****** *****.
**** ** ** ******* ** * search ******* ****** **** *******:
*** ******* ******* ********** ** *** performance ** *** *** *** ******** that *** ******* *** **** ****, the ***** **, *** ***** *****, and * ********* ***** ** *** users:
**** ** **** **** ***** ** just **** ******** ******* *** ** contexts *******. **** **** **** ****** at ***. *** ** *** *** 10 ******* ** ******** ** * 60 ****** *****, *** ****'** ***** to ** ********. ** *'** ***** with *** ******** ***** *** ***** possibly **, ***** ** *'** ****** for * ****** ***. *** **** it's ***** ***** *** ** ****** all ** ******, **'* ******* ** all *** ******* ***** ***** ** a ***. ** **'* ******* ** this ***** *** *** *********** **** the ******** **** ** ******* **** is *** ***** **** ****, *** video **, *** *** ********** *****. We **** **** ** ***** **** for *** ******* ******** *** *** developers *** **** **** ****** ** the *** ***** ***** ** *** search ******* ** **** **** *** use ** ** * ********* ** a *** ** ******* **** ***** to ***** ***** ** ***** ********.
*** ******* ***** *** ********* ** developed ***** ** ******** ******** *** objects, *** **** ** ***** ** trained *** *****' ******** *** **** scenarios *** ***********. **** ***** ** Twelve ****' *** *** ** **** for ****-**** ******, **** *** *** detailing *** **** ********* **** ********* of *** ********:
* ***** *** *********, ** * few ** **** **** **** ********* clever **** *** **** *** *** API. ****'** **** **** **** **-** second *****, ** **** ***** ***** up ***** ***** ********** ***** ** the *** *** **** *** ****** query **** ** *** ** **** a *********, ***** *****. ** ***** time *** *** * *** *** or *********, **** ** * ************. There *** ********* ****'* **** ** implementing ** **** ** ***** ***** our *** ** * **** **** if *** ****** * ** ** second *****, ** * **** **** detection... *** ** *** **** ****, I **** ** **** *********** **** this ****** **** ****. ******** *** at *** ******, *** **'** ********** looking **** **** **** ** ****.
*** *******'* ****** ********* *** ******** include *** ********* *** ********** ****:
** *** **** ******, ** *** actually ******** ** *********, * ****** other ********* **** ********** ***, *****, some ** *** **** ************* ********* that **** ** ***** **** ********** or ** ********* *** ******** *** embeddings. *** **** ********** ***, ** like **************, ************* ** **** *****, to ***** ****** ***** ***** **********. And *** ** **** *** ** powered ** * ****** ******* **** can ****** ********* ********** ***** ** once. ** ****, * *** **** to ****.
*******
*** *** ******** ******** ** *** for **** ** *** **** *** or *** ********* **** *** ******. Physical ******** ***** ************* **** ***** to *** *******, **** ** ** includes ******** **, ** ** **** be ********* ** *** **** **** of ******** ****** **** *** ******* or ******* **** ***** ******* * finished ******* ** *********** *** *** users.
****** **** *** ****** ** ********** $10 ******* **** ******, *****, *** Samsung, ******** *** ***** ****** ** $27 *******,********** *******.
***** **** *** ********** *******, ****** Labs ******** ********-**-**** ************ *****, *******-*, **** *** *** *********.