Transcriptly

The Journey from India to Building a $500M AI Video Product | Gaurav Misra, Captions

EO13:10Added Aug 31, 2026

Gaurav Misra is the co-founder and CEO of Captions, an AI video creation tool that's making it easier than ever for creators to make amazing content. Before starting Captions, Gaurav worked as a machine learning engineer at Microsoft and later switched things up as a designer at Snapchat. That mix of tech and design experience turned out to be the perfect recipe for building Captions, which is now based in New York, has raised $100 million in fun

Watch on YouTube →
Contributed by 刘嘉琪

Transcript

Transcript format

  1. Intro

  2. 00:00I grew up in India. I had no idea or no
  3. 00:02concept of like what is Boston? What is
  4. 00:04New York? I hadn't been to any of these
  5. 00:05places. Boston generally is a very
  6. 00:06multicultural environment. People come
  7. 00:08there from a lot of different countries
  8. 00:09and I went to Boston University where
  9. 00:11that's especially the case. After
  10. 00:13graduating from college, I actually
  11. 00:14started a PhD in machine learning which
  12. 00:16I actually dropped out of. I left my PhD

  13. How I Became an Entrepreneur at 16

  14. 00:18program and I joined a startup. People
  15. 00:20thought I was crazy that I would just
  16. 00:22drop out of it and get kind of go join
  17. 00:24some startup, right? But I think it was
  18. 00:25the right decision looking back at it.
  19. 00:27It was a very high-risisk decision, but
  20. 00:29it worked out. Especially in the US, and
  21. 00:30I think that's what makes the US such a
  22. 00:32special place. There's definitely so
  23. 00:34much opportunity that I think it's
  24. 00:36better to be a little bit more of a
  25. 00:38risk-taker. Most of the time, you can
  26. 00:39take a risk and actually get rewarded
  27. 00:41for it. I think people don't bias enough
  28. 00:44towards taking risks. If you just do
  29. 00:45what everybody else is doing, then
  30. 00:47you'll just end up wherever everybody
  31. 00:48else is going, right?
  32. 00:53My name is Goraf and I am the co-founder
  33. 00:55and CEO of Captions. Captions is a video
  34. 00:58creation company. Our goal is to make it
  35. 01:00easy for anybody anywhere to be able to
  36. 01:02tell their story through video without
  37. 01:04actually going through the physical
  38. 01:05process of recording or editing a video.
  39. 01:08We've raised over $und00 million in
  40. 01:10funding. It's gotten to tens of millions
  41. 01:12of users and over 100,000 plus users
  42. 01:16daily.
  43. 01:22I was in the US for the first four years
  44. 01:24of my life and my parents moved to
  45. 01:26India. I grew up in India. Moving back
  46. 01:28to the US like I had no idea or no
  47. 01:30concept of like what it would be like.
  48. 01:33What is Boston? What is New York? Like I
  49. 01:34hadn't been to any of these places. My
  50. 01:36parents were like Boston's great you
  51. 01:38should go there. And I was like sounds
  52. 01:39great. Let me do that. So at Boston
  53. 01:41University it was an interesting
  54. 01:42situation because coming from India sort
  55. 01:45of the curriculum there especially
  56. 01:47around computer science is just very
  57. 01:49advanced compared to the US and a lot of
  58. 01:51the things that we actually were being
  59. 01:53taught in college we had already covered
  60. 01:55in high school for the most part. A lot
  61. 01:56of college classes were essentially
  62. 01:58repeating what I had already done in
  63. 02:00high school. It probably was the case
  64. 02:02for a lot of people from India I
  65. 02:03imagine. I think that did give me a
  66. 02:05chance to actually go and explore
  67. 02:06outside of just that and kind of see
  68. 02:09what else I can learn and what more I
  69. 02:11can do. I did tons of projects. I like
  70. 02:13worked on projects with my roommates. I
  71. 02:15worked on a lot of different apps that
  72. 02:17we launched and this was like in 2012
  73. 02:192011 when the app store was still early
  74. 02:21in high school and going into college. I
  75. 02:23had started a kind of a company in a
  76. 02:25strange way. I say that because I
  77. 02:27actually didn't realize it was a company
  78. 02:29until much later. I was working on some
  79. 02:31games and there was some need around
  80. 02:33like security software for those games
  81. 02:34and I decided to make it and the game
  82. 02:36company decided to buy it for me not
  83. 02:38knowing that I was like 16 years old and
  84. 02:40I ended up selling it to a bunch of
  85. 02:42other companies as well. I didn't
  86. 02:43realize at that time that this is what a
  87. 02:44company is is like you make a product
  88. 02:46you sell it but much later I was like
  89. 02:48wow that's kind of what it was. It was a
  90. 02:50interesting sort of experience to go
  91. 02:51through because it kind of transitioned
  92. 02:53from generally games and gaming into
  93. 02:55sort of more general skill set. I really
  94. 02:58enjoy learning new things. How do you
  95. 03:00learn things quickly and how do you
  96. 03:02figure things out quickly? It's
  97. 03:04something that not a lot of people talk
  98. 03:05about, but it's like one of the most
  99. 03:07important skills in my mind. Go to a new
  100. 03:09domain, you know, go to a new area,
  101. 03:11something that you haven't done before,
  102. 03:13try for the first time, figure it out
  103. 03:14quickly, and be actually useful and
  104. 03:16productive in it is like it's a pretty
  105. 03:18cool skill and it's pretty fun one to
  106. 03:20work on because you get to try out a lot
  107. 03:21of different things. You'll get to learn
  108. 03:22a lot of different things and focus on
  109. 03:24that as a skill in itself. Being able to
  110. 03:26do all those different things and
  111. 03:27figuring those things out is really
  112. 03:29[Music]

  113. First Career in Microsoft and Snapchat

  114. 03:34important. Through the early part of my
  115. 03:36career, I was really focused on machine
  116. 03:38learning as a central aspect of what I
  117. 03:41was doing. At Microsoft, I joined them
  118. 03:43on this new team they were developing
  119. 03:45called Azure Machine Learning. It's like
  120. 03:47a machine learning platform they were
  121. 03:48building. And then at Snapchat, there
  122. 03:50was a bit of a difference. I got a
  123. 03:52chance to really get more on the product
  124. 03:54side and eventually ended up joining the
  125. 03:56design team which was like a huge shift
  126. 03:58from what I had done previously and the
  127. 04:01idea of like connecting technology to
  128. 04:02actual design and actual product made a
  129. 04:05lot of sense. So I ended up starting a
  130. 04:07team there called design engineering
  131. 04:09which was like how do you bring together
  132. 04:11great design and great engineering to
  133. 04:12build amazing products right? One of the
  134. 04:14things I learned at Snap, if you think
  135. 04:16about Snap as a company, I think Snap
  136. 04:18was one of the earliest companies to
  137. 04:20actually get into video for consumer use
  138. 04:22cases, especially short video. It did
  139. 04:24get me very immersed into the space of
  140. 04:28like video in general, especially video
  141. 04:29for communication, short videos, and all
  142. 04:31the challenges both technical and
  143. 04:33product that are related to that. But I
  144. 04:35think in general, one of the things that
  145. 04:37I noticed was essentially a shift in the
  146. 04:40way that people were communicating. A
  147. 04:41lot of people thought Tik Tok was just,
  148. 04:43oh, Tik Tok, that's like just people
  149. 04:44dancing, right? It's for fun. But I
  150. 04:47think the reality was that yes, there
  151. 04:49were a bunch of people dancing, but
  152. 04:50there was a lot of people talking and
  153. 04:52there were a lot of people telling
  154. 04:53stories and, you know, sometimes bad
  155. 04:55things that happened to them, good
  156. 04:56things that happened to them. People
  157. 04:57were like talking about ideas and like
  158. 04:59what could be done and what should be
  159. 05:01done and people were arguing with each
  160. 05:03other, right? And I think that was like
  161. 05:05a shift in the way people communicated
  162. 05:06from like that's the type of stuff that
  163. 05:08would happen like on Twitter or like on
  164. 05:10text beforehand, right? and it was now
  165. 05:11happening on video for the first time.
  166. 05:13So, I got pretty familiar with all the
  167. 05:16different ways that people are
  168. 05:17communicating and kind of what works
  169. 05:19better and what doesn't

  170. Success is in the Small Decisions

  171. 05:24work. When we started the company, we
  172. 05:27were thinking about building a social
  173. 05:28network and so we were thinking, okay,
  174. 05:30there's like a distribution part and
  175. 05:32then there's a creation part of the
  176. 05:33social network. We thought, hey, let's
  177. 05:35start with the creation first. So we
  178. 05:37started off with the idea of literally
  179. 05:39just adding captions to videos, text on
  180. 05:42videos based on what people are saying.
  181. 05:43The reason we did this is because we saw
  182. 05:45there was a big accessibility movement
  183. 05:47at that time. There were a lot of people
  184. 05:49sort of manually writing text on top of
  185. 05:52their videos and it was spent an hour
  186. 05:54like typing it all out. There wasn't
  187. 05:56actually a solution for this that was in
  188. 05:58the market at that time. So we were the
  189. 06:00first ones to build it and the absolute
  190. 06:02first project we built it in one weekend
  191. 06:05just as like a test essentially and we
  192. 06:08put it on the app store and the next day
  193. 06:10we woke up and it was like 600 people
  194. 06:13per minute using it basically it was
  195. 06:14completely at the top of the app store
  196. 06:16with no marketing no nothing done
  197. 06:19essentially and I think that was like
  198. 06:21the moment where it was like wow it
  199. 06:22can't be this easy right like it really
  200. 06:24can't be because we had been told right
  201. 06:26like it's going to be a lot of work to
  202. 06:28find product market fit and things like
  203. 06:29that. But I think that being said, we
  204. 06:32had just started the company and we had
  205. 06:33this idea of a social network and things
  206. 06:35like that. I think we were like, "Okay,
  207. 06:36this is working, but I wonder if we
  208. 06:38should go back and like really think
  209. 06:40about the social network aspect of
  210. 06:41this." And so we tried to figure out how
  211. 06:43to get people to post on the platform
  212. 06:45and things like that until we
  213. 06:47essentially rediscovered captions. And
  214. 06:49this was a interesting moment where we
  215. 06:51were working on another application
  216. 06:52which was called Spam App. It was like a
  217. 06:54photo sharing app. It was actually doing
  218. 06:56really well. While we were working on

  219. The Small Decisions he did

  220. 06:57that, we're looking at the bills and
  221. 06:59there was like a good $10,000 a month
  222. 07:01going to captions. You know, I remember
  223. 07:03talking to my co-founder and we were
  224. 07:04like, "Okay, let's just delete it
  225. 07:06because we don't want to spend $120,000
  226. 07:08a year on this, right?" We decided to
  227. 07:10delete it. I remember going back home
  228. 07:12and on Saturday going through like, you
  229. 07:16know, the UI to be like, "Okay, I'm
  230. 07:17going to press delete now." And while I
  231. 07:19was going to press delete, I was just
  232. 07:21thinking, "Ah, but like we worked so
  233. 07:24hard on this. Like I don't want to
  234. 07:25delete this, right?" Like it's like I
  235. 07:27want to keep it for memories at least,
  236. 07:28right? So I had an idea just in the
  237. 07:31moment quickly design a little payw wall
  238. 07:33where we ask people to pay for the app.
  239. 07:36If nobody pays for it then it's
  240. 07:37essentially shut down but we can
  241. 07:39continue using it because you know we
  242. 07:41work hard on it and if someone pays for
  243. 07:43it then it pays for itself. So we're not
  244. 07:45paying the bills basically. So I quickly
  245. 07:48made that in about 2 or 3 hours put it
  246. 07:50on the app store and launched it and
  247. 07:52then completely forgot about it. So, we
  248. 07:54continued working on spam app and it
  249. 07:56continued growing through the next, you
  250. 07:59know, four, five, six months. While we
  251. 08:02were doing that, I for some reason
  252. 08:04decided to check my account. And when I
  253. 08:06checked it, it was like at $500,000 or
  254. 08:09more and growing. And like the craziest
  255. 08:11part was not only did people pay for it,
  256. 08:13people retained on it. And it was
  257. 08:15growing in revenue completely on its own
  258. 08:17with like zero employees working on it.
  259. 08:20So, no employees. There were like 2,000
  260. 08:23open support tickets. Nobody was
  261. 08:25answering obviously, right? And it would
  262. 08:28continue to grow. You know, we had to
  263. 08:30have this like tough conversation of
  264. 08:32like, oh my god, like this is truly
  265. 08:34working, right? And it was tough because
  266. 08:37we already had another thing that was
  267. 08:39also working and that we were raising a
  268. 08:41new round on essentially, right? I think
  269. 08:43we discussed this for a couple of weeks
  270. 08:45and we ended up deciding we're going to
  271. 08:48switch to captions and we're going to
  272. 08:49just focus on creation as the primary.
  273. 08:52We switched over and literally the
  274. 08:55moment we launched the first set of
  275. 08:57features on top of the existing app.
  276. 08:59What was looking like a fast growth
  277. 09:02completely condensed because the new
  278. 09:04growth was exponential just like
  279. 09:07straight up. I mean, I think that's kind
  280. 09:09of the crazy thing that I think about at
  281. 09:10this point is like that split-second
  282. 09:13decision was the difference between two
  283. 09:15completely different paths. Here's the
  284. 09:16thing, like we don't know. The other
  285. 09:18side could have been even more
  286. 09:19successful. That's the tough part about
  287. 09:20these decisions is like you have to
  288. 09:22choose one and you'll never know what
  289. 09:24the other could have been. I think the
  290. 09:25big decisions are actually in the small
  291. 09:27decisions in a way and the small
  292. 09:29decisions are what actually you don't
  293. 09:31think about too much but actually turn
  294. 09:34the tide of like what will happen in the
  295. 09:36future. I think that was the case over
  296. 09:38here too, right? Because like what felt
  297. 09:39like a very small decision of like other
  298. 09:41app should we delete it or not seem like
  299. 09:43kind of obvious. Anybody we would have
  300. 09:44asked would have told us, yeah, just
  301. 09:45delete it, right? Whatever. But I think
  302. 09:47those small decisions actually are way
  303. 09:49more important than the big ones
  304. 09:51sometimes. And I think this shows quite
  305. 09:53a bit like if you look at any
  306. 09:54competition at a global scale, right,
  307. 09:56which building a startup is kind of a
  308. 09:58competition at a global scale. Like look
  309. 09:59at the Olympics for example, right? Like
  310. 10:01people are like within seconds of each
  311. 10:02other, right? Sometimes even
  312. 10:04milliseconds of each other. That's not
  313. 10:05because of some major decision they made
  314. 10:07while training or something like that.
  315. 10:09Everyone's very good, right? It's
  316. 10:10probably like some minor micro decision
  317. 10:12that happened in the moment like, you
  318. 10:14know, while they're swimming and they
  319. 10:16moved one arm at this angle versus the
  320. 10:18other angle. Like that's what's deciding
  321. 10:19who wins and loses, right? I think it's
  322. 10:21like those small decisions that count
  323. 10:22the most.

  324. How to Win with Quick MVPs and Feedback

  325. 10:29We did face challenges around like
  326. 10:30identifying the user and making sure
  327. 10:32we're building for the right user
  328. 10:33because when we launched the app like it
  329. 10:35was immediately sort of successful,
  330. 10:37right? And it immediately blew up. It
  331. 10:40made things easier and harder in some
  332. 10:42ways. I think one of the things that can
  333. 10:43happen when too many people join sort of
  334. 10:46your app or product very quickly is you
  335. 10:49kind of lose lose sight of who the
  336. 10:51customer is because there's a lot of
  337. 10:52people who are just checking it out.
  338. 10:53There's a lot of people who don't
  339. 10:55actually care about the use case, but
  340. 10:57they want to just see what's going on.
  341. 10:58Between all that noise, it's difficult
  342. 10:59to figure out like which one of these
  343. 11:01people are the actual user. Who's going
  344. 11:03to retain and for whom is this actually
  345. 11:05solving a real problem? That's really
  346. 11:06important because at the end of the day,
  347. 11:08without focusing on the right customer,
  348. 11:10you'll end up building the wrong
  349. 11:11product. Captions was a paid app from
  350. 11:13the beginning. And I think elements
  351. 11:14there that was interesting is like being
  352. 11:16paid actually helped filter our user
  353. 11:19base a little bit and to only those
  354. 11:21people who were willing to pay. Those
  355. 11:22people that were willing to pay were
  356. 11:23people who had real pain point. And so
  357. 11:26with that, we got really good feedback
  358. 11:28from users, right, around like what we
  359. 11:29should build and what we should
  360. 11:30prioritize and what they want from us in
  361. 11:32the future. We decided to keep the app
  362. 11:34paid right from the beginning all the
  363. 11:36way till today because of that reason. I
  364. 11:38think on the technical side, we try to
  365. 11:40avoid as much as possible technical
  366. 11:42challenges that it's sort of unproven
  367. 11:44essentially, right? Like I think one of
  368. 11:45the mistakes that can be made is like
  369. 11:46spending a lot of time building
  370. 11:48something very technically complex
  371. 11:49without fully recognizing whether
  372. 11:51there's product market fit there or not.
  373. 11:52We spend a year building something and
  374. 11:54nobody wants it. Right? That happens
  375. 11:55very often and it's happened to us too.
  376. 11:57The key factor there is like how do you
  377. 11:59reduce the technical risk to like an
  378. 12:01MVP? How do you test with like just the
  379. 12:03bare minimum? Get something out in front
  380. 12:04of people. Have them see it. Have them
  381. 12:06try it. If they don't retain, then don't
  382. 12:08invest more in it, right? Try something
  383. 12:09else. And I think that's the other thing
  384. 12:11is like not going deeper and deeper into
  385. 12:13the rabbit hole almost in a way. I think
  386. 12:15there's also tendencies of like building
  387. 12:17the first version seeing that it doesn't
  388. 12:19completely work and then justifying it
  389. 12:22by saying well if it only had this
  390. 12:23feature and that feature and that third
  391. 12:25feature then maybe it would have worked
  392. 12:27right but I think the reality is like it
  393. 12:29would not probably not have worked with
  394. 12:30as many features as you want. It would
  395. 12:32only have worked with a single feature
  396. 12:34only one of those things right whatever
  397. 12:36actually people want at the end of the
  398. 12:37day. So, I think the better skill is to
  399. 12:39just learn how to quickly build stuff.
  400. 12:41Try as many things as possible. The most
  401. 12:43important thing is getting stuff out in
  402. 12:45front of people, testing it. If it
  403. 12:46doesn't work, kill it and start again.
  404. 12:48And keep doing that as quickly as
  405. 12:50possible.
  406. 12:58[Music]
  407. 13:03[Music]