#266 CaptionHub CEO Tom Bridges on AI-Powered Real-Time Media Accessibility

0:00Tom Bridges:CaptionHub does not and is not and will never be a service provider in that respect. What we can do is we can provide hopefully really clear and intuitive tools that people can use on a day-to-day basis in order to do the work they need.

0:20Hey everyone and welcome to finally a guest episode again at SlaterPod here today on the podcast. We welcome Tom Bridges. Tom is the founder and CEO of CaptionHub, an AI-powered collaborative platform for multilingual subtitling, voiceover, and on-screen text localization. Hi, Tom, and thanks so much for joining today.

0:39Tom Bridges:Hi, Florian. Thank you very much for having me here. What part of the world, what city, what country are you recording this from today? I'm currently in Bath in the southwest of the UK. Southwest of the UK. Last time we met was in the Silicon Valley. It was at SlaterCon, yes. Yeah, that's right. That was a great event. Thank you. So now we're back in our respective part of the world again. So tell us a bit more about your professional background. Like you shifted originally from a post-production role to multimedia localization. Just tell us a bit of the background, set the scene for the listeners and viewers here.

1:13Absolutely.

1:15Tom Bridges:Well, I started off my career in architectural visualization, as it happens. And that then flowed through into a post-production company, which then flowed into a post-production company. And after the post-production company, it was then motion graphics and visual effects. And it was part of that motion graphics and visual effects journey, in a particular motion graphics side of things, that led me into the subtitling world. Which was kind of by accident. if I'm being entirely honest, but we were exposed to the problem set of a very large customer who required a lot of localization in a very short amount of time.

1:53Tom Bridges:And we ended up developing a tool, and we licensed that tool to the customer, and that's how CaptionUp was born in the first place. Interesting. So the original problem we were trying to solve was big, big customer had kind of a left-field problem for you at the time, and you're like, okay, this could be something I can scale with many, many other customers as well. Precisely. It was localizing a five-minute video into, I think, 16 different languages, but managing the entire process overnight. So that was transcription, translation, the video was changing, there were lots of different stakeholders in multiple parts of the sign-off, and everything was going back and forth with Excel spreadsheets.

2:38Tom Bridges:So it was kind of crazy and um it the the workflow almost sort of really offended my sensibilities it was just so broken um and there's got to be an easier way there's got to be an easier way and the obvious solution um was to have you know a single source of truth and to gate access to that and make it secure and you know um you know allow the right people to have the right access at the right time and it was like the people doing the translation had never seen the original video for instance So it was just sort of mind-blowing how this kind of process had actually come into existence in the first place.

3:19Tom Bridges:But yeah, we developed a tool and that tool became CaptionUp. The spreadsheet is the backbone of the localization industry, has been. And yeah, it has offended my sense of productivity many times in the past also. The spreadsheet is a great, great tool, but it just doesn't work when you add in this concept of time, right? The whole point about multimedia is you've got an extra element, and there are lots of things that are very subtitle-specific, but that time thing, a spreadsheet has no sense of that. But originally, do you have any experience on the developer side or tech side, SaaS side, or did you just say, well, I'm the business guy, let me bring in the developers here?

4:03Because obviously, it's a technology, right?

4:05Tom Bridges:Yeah, no, absolutely none. So I had an idea. I'm relatively technical, so I had an idea of the kind of thing that was possible. But no, I'm not a coder and I can't code. So I had an idea, hired a developer, showed our customer.

4:28Tom Bridges:This developer worked for a month. We came up with a proof of concept, showed it to the customer, and the customer loved it. And we ended up licensing it to them nine months later. And I thought, wow, software's easy. Have an idea. Hire a developer and license it to one of the biggest brands in the world. And I've spent the last decade finding out quite how lucky we were with that initial set of circumstances. But, you know, we've gone from that initial sort of point of solving a very specific problem for a very specific customer, you know, based around subtitle localization, you know, and we're now 10 years on with a wider market view.

5:13Tell us a bit about that wider market view in terms of kind of service mix, client segments, geographic footprint, like where's the team, where are the clients based typically?

5:23Tom Bridges:So the wider market view, you know, we realized early on that, you know, we had a product that was based in the cloud. And, you know, there are some much, much smarter people than we are doing amazing things. And so even back in 2016, we were integrating with speech recognition and machine translation. And obviously in recent years, people are doing some amazing things with AI. you know we we realized that that connectivity and integrating with other tools was a really important thing and then of course you know the the human in the loop that's a that's a critical part and something that that you know is is very very important and so you know so we've gone from this point solution to being something that is a true sort of end-to-end multimedia localization solution.

6:12Tom Bridges:So way beyond subtitles, synthetic voice, live, automating this, and all of the further applications of AI. With respect to the market that we're playing in, it's a truly global market, really. We have customers from all different geographies and a whole load of different customer sectors. The bulk of our customer profile is still in the large enterprise. But of course, we're doing more and more in broadcast as well. And LSIs are an important market for us. I want to talk to you about all of these individual components. Also, of course, AI, the enterprise section, also the kind of rise of YouTube creators as potentially big clients.

7:07But first, I guess you also went to the IBC conference in Amsterdam like every year. It's a little bigger than SlaterCon. I wish maybe at some point we were IBC sized. But tell us a bit about the vibe, the mood, like the tech that was there. That would be relevant for people in the language industry.

7:26Tom Bridges:Yeah, as you say, it's on a different scale. It's not quite as big as some of the really big shows in Vegas, but it's definitely up there. the floor space you can measure in acres really. And AI, as you can imagine, is a really hot topic. We were in the AI tech zone and we had a huge stand this year. It was bigger than my first plat. And it was a great event. We shipped out the whole team. And there's just a palpable sense of excitement about the possibilities and what's out there. And I think IBC also, just by virtue of its location in Amsterdam, it's a natural sort of focal point for more sort of international markets, which obviously plays to what we're offering in terms of multimedia localization.

8:21Did you launch Timbra, that's a real-time media localization suite, just before IBC? And that was kind of the launch pad, as it were, for that solution?

8:30Tom Bridges:Yeah, we were introducing Timber as a concept to the world. In lots of ways, really, it's a repackaging of some of our existing technologies. When you say live multimedia localization now, there are a whole load of different applications there. So we started out with something that was purely based around web broadcasting. We've got some remarkable technology for HLS delivery that is kind of magical. You've got video-on-demand style subtitling, but on a live video stream. But that's expanded to be, you know, more and more, to be a wider offering. So we've got an in-room product, and, you know, there are various other things we're working on.

9:18Tom Bridges:So Timbro is the encapsulation of lots of these different related technologies, and that's what we are introducing there. But it's not, you know, what we're announcing wasn't sort of new per se. You know, lots of what we do is in use by some huge organizations. So Timber has provided the translated subtitles for Amazon's reInvent or AWS's reInvent conference. That happens on a yearly basis. Oracle's Cloud World, Intel's Foundry events, you know, device launches. So it's already sort of highly production proven, but it was really nice to be able to stand up on stage at IBC. One of our partners and end user, the company called Coribium, their CEO stood up alongside and gave the demo, which was really cool.

10:19Tom Bridges:And yeah, it was really well received and it was a privilege to be able to launch at IBC. So, you mentioned there AWS and what was the other company you mentioned? Also, big enterprise. So, Oracle. Oracle, all right. So, AWS and Oracle. So, basically, I was going to ask you that CaptionUp is obviously full enterprise grade, which is just kind of proven by mentioning these giant companies, right? But you're competing in like a super busy market. It's gotten way, way busier than let's say seven or eight years ago, right? with hundreds of maybe lower end products, right? People are just kind of putting things together on top of some AI and like, you know, basically have a slick marketing pitch.

11:04How do you differentiate currently, like having something that's super, super deep against much shallower kind of potential drivers in the lead gen process, not once you get in front of the customer and like RFPs and everything, but maybe at the top end of the funnel.

11:19Tom Bridges:It's a really interesting question. I mean, you know, and I guess if a customer is looking for a pure point solution, then, you know, in lots of ways, we don't really compete. You know, as you've said, you know, we play in the enterprise space. And, you know, we offer a full spectrum of tools, you know, across that multimedia localization, you know, problem set. and one of the great advantages about being AI agnostic, which we are, is that we can integrate into the people who are creating these amazing foundational models and we can offer our customers a curated but best of breed choice of technology that they use inside of their workflows.

12:12Tom Bridges:So, as I said, if a customer is after a sort of pure point solution, then we don't really tend to compete in that level. If they're after a platform to manage the entirety of their multimedia localization operation, there's really not a great deal of choice that they have out there. And the enterprise, of course, has got very different needs to the consumer or prosumer space, particularly around security, in terms of audit trails, in terms of everything else, the custom contracting, for instance. We offer all of that and all of the various certifications. So I hope that answers the question. It did.

12:59It did. It did. Let's go a little bit, start touching on AI. So you guys are offering obviously AI speech recognition, AI translation, synthetic voice, but you also mentioned before even like the kind of expert in the loop component, expert in the loop workflows. Now, who offers that? Are you guys offering this? Are you basically plugging in third parties and you're the tech provider only or how does that work?

13:21Tom Bridges:We're the platform, you know, we're the glue, if you like, that holds it all together. Caption Hub does not and is not and will never be a service provider in that respect. What we can do is we can provide, hopefully, really clear and intuitive tools that people can use on a day-to-day basis in order to do the work they need. And how much human-in-the-loop involvement is entirely up to our customers and what shape that takes is entirely up to our customers. So if our customers want, if they've got sort of small teams who they want to direct towards a particular video, they can do that. But they can also go up to having a multi-vendor strategy inside of Caption Hub.

14:10Tom Bridges:So you can have different organizations sign off different stages of videos. And of course, everything in between. One of our longstanding customers is TED. Ted have over 70 ,000 users on the Caption Hub platform, and they're using that in an entirely sort of self-managing way, you know, to manage a multi-stage, you know, approval process. So there are lots of different ways of managing the human in the loop. You know, Caption Hub, at Caption Hub, you know, we leave that down to our customers, but we provide the tools to be able to make this happen. And I think one of the interesting things as well is, you know, I talked about our three pillars earlier on, you know, the AI, the connectivity, and those human in the loop processes.

15:01Tom Bridges:I think one of the unique strengths that we have at CaptionHub is the ability to automate between all of those things. So, you know, with a single command, you can send a video to CaptionHub. You can have it transcribed with, you know, your own glossaries, your own custom dictionaries. you can then route that to your internal team or through multiple vendor steps. And then once you've got that back, on approval, you can then generate translations. You can generate translations with different machine translation providers, different term bases. Again, sort of optimizing the AI. But again, if you need the human and loop processes, you can do that too.

15:41Tom Bridges:And then once those have been approved, then you can automate the delivery of the final product. And whether that's, I don't know, voiceover tracks going to a video platform or whether that's burnt in subtitles going to an S3 bucket or whether that's captions being published to YouTube. You know, there are all sorts of different things you can do and you can do that with a single API call. And that's massively powerful and it's really massively scalable as well. Have you seen a dramatic shift in terms of maybe less demand for human in the loop in the past, let's say, 18 months? or is it kind of steady or does it correlate with volume?

16:19Tom Bridges:I think it's a really interesting question. And I think it very much depends on the content type.

16:32Tom Bridges:And I think there are some kinds of content where having an entirely automated process is just fine and completely appropriate. So one of our customers a few years ago wanted us to localize their back catalog of videos. Now, someone else had done all of the transcription. So we were able just, you know, we're spending five minutes in CaptionHub to set up an automation that when they tagged their videos in the video platform, they got ingested into CaptionHub. and CaptionHub ingested those videos, pulled in the transcription that was already there, translated it, true sentence level translation, leveraging best of breed AI, and then just sent captions back, sent translated subtitles back to the video platform.

17:28Tom Bridges:And they had 3 ,500 videos. They added a tag to 3 ,500 videos. 48 hours later, they had quarter of a million minutes of subtitling done. And that's an entirely automated way. And for some kinds of content, that's absolutely fine. For other kinds of content, tier one marketing, you'll always need that human in the loop process. Let's talk about a couple of kind of segments, market segments. Maybe I want to talk about YouTube and the creator side, then maybe major kind of feature films and then move on to the broadcasting space where you're super strong in, right? So what are your thoughts about YouTube now allowing big creators or even smaller creators like Slater?

18:12We could now upload our multilingual audio tracks. Do you think that's kind of a big unlock for AI dubbing or is it maybe a little bit less of a big deal? I just keep thinking it's a huge deal and it should unlock a ton of latent demand, as it were.

18:26Tom Bridges:I think it's really interesting. And I think, you know, the fact that companies like Google are offering this is, you know, shows that there's demand out there. And, you know, and perhaps it's demand that a lot of people didn't know they had until they, you know, they had the tools put in front of them. You know, and in terms of, you know, we're not trying to compete with YouTube. You know, like auto captions have been around on YouTube for a really long time. You know, they're fine. I haven't done too much. I haven't had too much experience with the synthetic dubs in YouTube, but I'm sure to a certain level, it's fine.

19:08I think where CaptionHub comes in is when people need that last bit of control

19:19Tom Bridges:and all of the other things that I've talked about earlier in terms of what an enterprise might need. but by the same process by the same token if what you get out of YouTube is good for your needs then honestly I think that's something to be celebrated I mean yeah the DI dubs that YouTube has kind of integrated their own solutions really bad I can tell you that but now they're so bad that they must have accelerated the launch of basically you being able to upload your own dubs and that's what's there now right so yes exactly yeah that's available now And that's something we'll be integrating with quite soon, the ability to send a caption of voiceover to YouTube.

20:01Tom Bridges:We already have a YouTube connector in place. Yeah, that should be a pretty big on-ramp. Are you guys at all in the kind of major feature film space? Like, I don't know, the Hollywood, the world of Hollywood and related big, big feature films? Not so much, no. Some of our customers do use us for episodic content. you know we're not used for future films at this point in time the you know the the enterprise market and the broadcast market have very very different needs honestly where we're focusing most of our efforts for the broadcast market despite the fact that Caption Hub is a fully frame accurate platform you know can absolutely be used for episodic TV but where we're seeing most traction at the moment is in our live subtitling offerings with Timber.

20:50It's so interesting, the live space. So the broadcasting and live market, what's kind of the hardest technical challenge? I mean, live seems just generally very hard.

21:01Tom Bridges:It is, and it's hard in sort of in different ways as well. You know, not just from a technical level, but at an organizational level, being able to, you know, provide the level of support that large organizations need. You know, they've invested tens of millions of dollars and their events sometimes. And, you know, nothing is left to chance. So it's a very difficult problem to understand and therefore to solve. And I think part of the issue with Live is that there are a whole load of different use cases for it. You know, as I sort of alluded to earlier, you know, I was saying Timber is a collection of tools.

21:47Tom Bridges:in lots of ways. There are lots of different ways of people generating and consuming live video and live audio for that matter as well. I think something that sort of stood us in really good stead though is that appreciation of, you know, what good captioning looks like. At Caption Hub, we've got a bunch of algorithms, for instance, that take, you know, if you were to just upload a video, a video on demand asset ticket to the Caption Hub platform and transcribe, We send audio off to a speech recognition provider, and then we get a big transcript back from that speech recognition provider, and then turn that into captions that a human would have created.

22:31Tom Bridges:And this is core technology to us, and it's kind of much harder than it sounds. You've got to pay attention to where cuts are in the footage. You've got to pay attention to grammar, to subtitle constraints, so, you know, characters, character limits, reading speed, that kind of thing. And so we'd gone to a great deal of trouble to develop this for our video on demand side of things. But we can reuse that, all of that R &D, you know, for life as well. And so that's what I was talking about with Timber and the RHLS product. you're getting genuinely video on demand quality captioning, but in a live sense.