#257 The 50 Top Language AI Startups of 2025

0:00Florian:50 of the most innovative language AI companies founded in the last 50 months.

0:09Florian:All right, we're back. Hi Esther. Hi Florian. We are back because lots to talk about. We just launched our 50 under 50, the Slater Language AI 50 under 50. It's the third edition. So we've over the past two years dug out 150 super interesting startups and we want to talk about this today, but first you do want to go and register for SlatorCon Silicon Valley happening in September or on September 4th in Menlo Park. The agenda is nearing completion. Really, really cool. Looking forward to being back in California in less than two months. A lot of the team is going there and then Alex Edwards, senior research analyst is traveling on to New Orleans to join the ALC summit, which is happening, I think, about 10 days after SlaterCon.

1:02Florian:Very nice. Exciting US trips we have coming up. So go there, register, and meet us in California. On the agenda today, again, we want to talk about the Slater Language AI 50 Under 50. Then you're going to talk a little bit about demand in the public sector coming out of France and the United States, we do want to talk a little bit more about Propra and Suricom or Propra buying Suricom. We spoke about this just before we started recording. I think it's a very important deal and it kind of was breaking news last week. So we want to talk a bit more about that. And then IPO teasers, IPO language AI, IPO announcements, teasers, 11 Labs just said something interesting so I want to comment on that.

1:51Florian:And then finally, geek out a little bit on the limitations of reasoning models for AI translations. And there are 10 of them, according to our very own Maria Stassi-Mioti, who unpacked that this week. All right. 50 Under 50 published a few hours ago at the time of recording. So what is the 50 Under 50? What's the 50 referring to, Esther? Well, which 50? 50 companies. Yes. It speaks for itself and under 50, meaning under less than 50 months old. So, approximately four years since they were founded slash launched. Yeah. Correct. So, it's not 50 companies with founders under 50? No. Although, I mean, probably if you surveyed them, they might be under 50, but we do not know.

2:41They might be under 50.

2:42Florian:You know, the elder generation. All right. So 50 of the most innovative language AI companies founded in the last 50 months. So we've worked quite hard on this, I have to say. Not easy. We did quite a large market scan, reviewed hundreds and hundreds of companies, both LTPs, Language Technology Platforms, and LSIs, Language Solutions Integrators. Criteria, again, was they're going to be under 50 months. They're going to be standalone, so not owned by some parent company, and they have to enable language conversion or multilingual language generation. Right? All right. Scanned the market, hundreds of companies, went on to the selection, long list, I think over a hundred, and then started pinging the CEOs, the founders to understand a bit more about the company where possible.

3:37Florian:We actually tested out some of the free versions of the tools. So this was great. And then what we were looking for really was kind of innovation, growth, product that solve actual buyer problems, have a unique positioning, a unique go-to-market strategy, have strong use cases, have already built up some use, some case studies and can share some client logos. Great. All right. So then we selected 50, grouped them into five categories, which represent the current hotspots of language industry activity. Maybe Esther, you can walk us through those five categories. Yeah, so category one, we have multilingual video and audio.

4:17Category two, live speech translation. Category three, transcription and captions. Category four was translation and text generation. And the final category five was accessibility.

4:29Florian:Those are our areas. Yeah, accessibility is something that we're now really adding much more to the core area of coverage that we're looking at. In terms of the challenges, spoke to the research team. So they were telling me that there's still a lot of LLM wrappers that are kind of flooding the market, particularly in AI dubbing and text-to-speech. So we needed to sort the wheat from the chaff. Also, one challenge was standalone versus subsidiary. A lot of products that on first glance you think are standalone actually belong to big tech companies. I was told particularly in Asia or other types of kind of broader tech conglomerates.

5:11Florian:So this is, yeah, interesting. It should probably explore this separately. And then, yeah, just generally there were just so many strong contenders. Apparently the code of long list of companies was a lot stronger this year than any previous year, which indicates that the market's really maturing fast and there's so many new entrants coming in. It's just so interesting. You have this kind of two-tiered market, right? On the one hand, you have the more established LSIs being generally more on the worried side, right? AI coming in, what do we do? And then the other side, you have this kind of incredible activity in startup land, which is obviously much more in a hopefully positive frame of mind because they started a company and they're hoping to get traction.

5:59Florian:So, yeah, and we're trying to bridge the two. So, bring some maybe much needed market kind of realism to some of the startups, but also signaling to the established player that, hey, there's a lot more cool stuff you can do and it's not just all going to go to open AI. Yeah, I think it's like realism on both sides, isn't it? So it's like, hey, there's a lot of hype. This is the reality. And then, hey, there's a lot of sort of worry and concern. You know, let us help you move towards a slightly more optimistic, proactive view. Correct. So let me just for those on YouTube, briefly spin up the logo map we put together, which is also available on the article.

6:45Florian:So yeah, here you see for those that are looking at this on YouTube, the big category there is multilingual video and audio, a bunch of logos there on the logo map, live speech translation, translation and text generation, transcription and captions and accessibility. So I'll lift this up for a sec for those that are on YouTube. So again, the main theme this year was like you need to have the whole product experience. There needs to be some verticalization, specialization. Some of the segments are maturing. And so, yeah, this year startups couldn't be further from those generic tools, maybe prompt-based, just LLM wrappers that we looked in the past.

7:27Florian:I said before there's many out there, so we tried to really select those that are not. But this year's tools that we selected really offer buyer control, integrated workflow, have very kind of business specific capabilities. And again, that was the key here that solve real buyer problems, right? Buyers and users, they know there's a ton of cool stuff out there, but you know, you sign up and then you get disappointed because the thing box out on you within seconds. So we try to really select companies that solve real buyer problems. All right, let me unshare for those on YouTube again. So let's start with, I just want to maybe highlight a couple of companies.

8:09Florian:We can't go through all the 50 because otherwise people will drop off the podcast. But so maybe let's just mention a couple that and absolutely not, you know, those that are not highlighted now are great too. You're all great. Cool. So multilingual audio, for example, one is aunion.ai, which targets high fidelity automated dubbing for professional content workflows, a very B2B, Linguana is super interesting and we talked about YouTube quite a lot. I hope nobody got upset at YouTube for my rant last week, but this is an interesting one because Linguana, they provide fully managed localized YouTube channels at no cost, but they take a cut of your revenue.

8:57I saw this one recently because I think we spoke about them in the funding section. We must have.

9:03Florian:Yeah. Yeah. Yeah, but it's a very interesting model in terms of what they're offering, the management and the localization, doing that whole piece. They take a cut. So the better they localize, the more traffic you get, the more views you get, the more ad revenue you get from YouTube, and they take a cut. Yeah, so incentives are extremely well aligned. And I love the, you know, how long has the industry been talking about? Okay, localization is a revenue driver. I mean, here you have, I mean, it couldn't be more incentivized to produce ROI. I think if you're going to try, I think if you're going to implement this model somewhere, I mean, YouTube is the place to do it.

9:44YouTube is the place to do it.

9:46Florian:Yeah. And then on live speech translation, very interesting. So we had Jade Health on the podcast, you know, maybe half a year ago. And so they were featured, Palabra AI is, you know, they enable language Solutions Integrator to scale and productize live speech translation. We had them at SlaterCon London as well. So, very interesting products coming in live speech translation. Obviously, a super tricky problem. Many devil in the detail problems there that you need to solve. In translation text generation, we have Gridley who was also on the podcast, more on the kind of CMS side, combines localization and CMS.

10:26Florian:And then Lingo.dev. Did we speak about Lingo.dev in the past? I don't recall, but for some reason I really like the name. So... Lingo.dev. Yeah. They kind of jumped out at me for some reason. I think they're going very deep in kind of the, yeah, the development cycle already kind of enabling localization at the very kind of foundation. So very interesting. Go and check out Lingo.dev. Bloxweaver provides, you know, end-to-end content services supporting multilingual formats, model fine-tuning. I think very early stage, Bloxweaver with an X. So go and check that out. Transcription and Captions, typewriters.ai.

11:08Florian:That's interesting. That's a good name. So go check that out. And in accessibility, we have Synapse, which we had on our startup panel in London. Yeah. Another company called SignBridge, AI tool for real-time, bi-directional sign language translation. I mean, the whole accessibility space is such a fascinating one because there's so much upside in that. Think about it. I mean, with sign language, if that's actually solved, there's so much upside. This is a problem that so many people have. And it really makes life fundamentally easier if that's something you could just plug into any type of channel, right?

11:52Florian:Definitely. All right. So, look, we could go on and just go through all of these companies. But I think, you know, make sure to head to the website, 50 Under 50. It's out now, freely accessible. And congratulations to all the risk takers and entrepreneurs that are, you know, building here in this incredibly fast moving environment. So, yeah. Yeah. Proud of this list. Good piece of paper. Now, from startups and tech, we go to Emmanuel Macron, who has been in your part of the world. Yeah, he's been visiting. Apparently speaking too much French. Oh, really? That's the pushback I saw from all the ex-posts.

12:38Oh, you can never speak too much French. You can never speak.

12:42Florian:No, his English is excellent, right? So, I mean, he's a former investment banker or something like that. Anyway, so now that he can use Diplo. My seg was going to be one company, one product that might have made it onto the 50 under 50, were it not part of the government, i.e. not standalone, was Diplo. And I don't actually know how to say this, but it's Diplo IA. and I guess the IA being like intelligence artificielle, like the French version of AI, right? So DiploIA, DiploIA, however you say it, it is a new language AI tool that has been quietly developed and deployed as we reported by the French government for diplomats.

13:32So hence the Diplo in the DiploIA name. so this is a set of language AI tools like I said it was developed by the government's digital directorate the tools are designed to help these diplomatic agents of which apparently there are 13 ,000 when they're doing sensitive missions so whether that's at home or abroad they say so this stealth project as it was apparently very difficult to find sort of any research or information online, but it was actually sort of covered and discussed in the French press. French news site Acta Public, AP, talked a little bit about Deployer, saying it's an AI-powered system focused on multilingual translation transcription and that it directly addresses the specific needs of diplomatic field operations.

14:26So there you go, you've got your use case right there, that's your bio use case. And it's been in use since a few months now, since May. So yeah, pretty interesting that this has been going on within the digital directorate, the ministry in France.

14:44Florian:Of course, the French need their own tool. Of course. Specifically tailored to their very specific needs. Yeah, it's all very James Bond. Similar public sector is in the US. So SOSI or SOCI, I don't actually know how to really pronounce it. I normally say SOCI, but then the name is SOS International. SOS International. Your guess is as good as mine. So they've got a quarter billion dollar contract. Exactly. Yeah. So, I mean, more from the public sector. I think generally it's always good to check in and see what's going on. We do that via regular coverage on the website as well about some of the contract awards and contract or tenders that are up for grabs in various parts of the world.

15:32So here we have news that SoC was awarded pretty much a quarter of a million US dollars. I always get this wrong every single week. I need to upgrade my...

15:45Florian:We're in the billions. My thinking. Yeah, yeah. See, I'm just dealing low level in the millions. But yeah, I need to start thinking in terms of billions. There we go. Yeah, it's inflation, isn't it? Well, anyway, a language support contract has been won by SOCI. It's big. And so they have secured a language services contract from the DEA, so Drug Enforcement Agency, that involves real-time monitoring, advanced analysis of intercepted communications. So again, all very sort of mysterious and interesting within the public sector there. So yeah, huge contracts. I mean, SOCI has been working with the DEA and obviously other sort of federal agencies there for a long time.

16:32The company was founded back in 1989. So we're talking, you know, a lot of years of partnership between sort of government agencies and SOCI. In the past, I think five and a half years alone, they've been awarded 275 federal awards. So this is definitely sort of an area of specialism for them. They're not only focusing on language services as a company, but they're also looking at technology, logistics, human services and intelligence. So very much providing a wealth of services to the government agencies. Language being a subset of the intelligence category, which obviously is a nice place to sit under intelligence.

17:19so yeah they've also worked on hundreds of language services contracts including one with the daa from 2018 but yeah super super interesting that obviously you know we've got that kind of level of award being made within language services there and not just soci necessarily but we also have a heads up about another tender that is saying that the US Defense Health Agency is looking for LSIs. So I don't think we have specific confirmation of how much the contract or contracts would involve. But I mean, in terms of volumes, we're talking lots and lots and lots of hours and minutes in this time for interpreting.

18:11So this is a heads up about demand for healthcare interpreting in particular. They're looking for providers of translation interpreting within Maryland and Virginia. They say that the Defence Health Agency is looking for 24-7, 365 coverage in common, uncommon, as well as exotic languages, as they describe it, and ASL. They want a super fast call connect time for OPI, over the phone interpreting, of 60 seconds for Spanish and 120 seconds, i.e. two minutes for other languages. For the on-site interpreting piece, I think the turnaround time or time to be on-site is within two business days, and they're only considering certified interpreters.

19:03So lots of...

19:04Florian:Yeah.

19:08Florian:I'm a certified interpreter for an exotic language. Yeah, I think also the idea of certifications is, as we know, I mean, there's lots of different certifications. I don't know how specific they're going to be around that. And that's always a challenge, certainly here in the UK anyway. But yeah, in terms of volumes, we're talking hundreds of thousands of interpreting minutes and hundreds of thousands of words for translation which I mean that's sort of table stakes I suppose but this just by way of an example in terms of the sort of the size of the demand here we've got one medical center and its affiliates would be looking for 14 ,000 minutes of OPI 160 hours of on-site interpreting and 20 hours of VRI as part of that.

19:55So it is actually a one-year contract starting from October this year, 2025, but there are four additional option years. So that's sort of the extra extension. And the estimation is that if the contract runs for a full five years, there would be something like 6 ,000 hours of over-the-phone interpreting, more than 9 ,000 hours of on-site interpreting, and more than half a million or billion. That's a million, that's a million. That's not a lot. Anyway, hey, words for translation. But it's just, yeah, it's interesting to see sort of the volumes that we're talking about and also still sort of the criteria, the specification, the level of service, response time, all of that, which is so heavily dictated in some of these settings.

20:46Florian:I'm sure one company that, and I'm doing a segue here, that's bidding for this could be Propio, right? Now that it successfully acquired Suricom in a giant deal. Tell us more again. Let's just unpack this a little bit because last time it was kind of breaking news. So obviously this is, is it the biggest acquisition this year? Must be. Top of my head, yes. For sure. I mean, there are some other significant ones, obviously, that have happened in the first half. But I think in terms of probably, at least in terms of, let's say pure revenues, for example, this one probably is so far the largest. Yeah, I mean, just as a reminder, we covered this last week, but we're talking about Propio, which itself has scaled rapidly in the past couple of years through a recent M &A spree, in addition to organic growth, I'm sure, but buying in 2024, ULG, Acorbi, ASL Services, so some large companies as well that they acquired back in 2024.

21:45And then the news that broke last week was the acquisition of Siracom, which itself was already a leading interpreting provider. Specifically, remote simultaneous interpreting was sort of their sweet spot for healthcare. So, yeah, interpreting healthcare there, as well as for legal and public sector.

22:06Florian:They have these contact centers, or they used to have big actual buildings with interpreters in it, So, which is different from purely remote setups. But that's, I mean, my knowledge here is a little dated. Maybe they've transitioned, but I remember like three, four, five years ago, they had these contact centers. I think they probably still do. I mean, there was, I think reading in the press release, they were talking about having contact centers, or at least a presence, not just in the US, but in a couple of other countries south of the border as well. But yeah, this is definitely bringing together two companies that are already leading, particularly in the interpreting space, as already sort of standalone and obviously combined is a pretty major deal.

22:56In terms of revenues, they're going to be more than half a billion. Yes, I got that right. And I think we covered this in the article, so do go check it out. But Marco sees the CEO is talking about sort of this being a bet on healthcare and in terms of being a growth market and sustainable there.

23:15Florian:So who knows, maybe they're going to go public at some point and then we can get a lot more information. But they haven't said anything. Unlike the CEO of 11labs, Mati Staniszewski, who was talking to CNBC about that their eventual aim is to get the company ready for an IPO in the next five years. So, okay. So you're a startup tech, two to three years old, and now you're talking about going public in the next five years. It's like, okay, interesting. And then CNBC makes kind of a headline. The headline is AI voice startup, 11 Labs pushes global expansion as it gears up for an IPO. But you're not really gearing up for an IPO.