#288 The Language AI Startups to Watch in 2026

Speaker A: Language has really packaged inside solutions designed about much broader business objectives. All right, we're here with the definitely hottest episode of Slated Pot so far in the 288 podcasts we've recorded. Hi, Esther.

Speaker B: Hello.

Speaker A: How are you holding up? Good.

Speaker B: Yes, it's balmy.

Speaker A: Humid and hot, Humid and hot. We're in Europe and the heat is on. So for those of you who don't live, uh, in Europe, it's like, like 37 degrees Celsius here in Zurich, which is insane. So that's like nearly 100 Fahrenheit, I guess in London, it's marginally cooler, maybe 2 degrees.

Speaker B: Yeah.

Speaker A: Okay. And we're talking about it not because we like to talk about the weather, because we're recording this podcast in a place where there's no air con, because, as we learn now, there's basically no aircon in the countries that we inhabit. And, uh, yeah, that makes it challenging. But we're still here and we're still recording a podcast for you. Uh, probably things are a little bit cooler in San Francisco, where we are going in, uh, uh, the first week of September. So Wednesday, September 2nd and September 3rd, we're going to be in San Francisco at the phenomenal St. Regis Hotel, uh, in San Francisco. So, uh, we're super happy to have confirmed a couple of outstanding, um, uh, panelists for the investment panel that you're going to run. Esther, uh, from Khosla Ventures, one of the world's leading VC firms. Canoe Gulati is going to join us there, and Radhika Malik is going to join us. She's a partner at Dell Technology Capital. And so, yeah, we're very much looking forward to that. Two great additions to, um, uh, the speaker roster there. We also have Georg El, uh, from Frase, we have Marco Ciamperi from Translated, Gabriel Fairman from Works, Uh, then Teresa Marshall from Salesforce Link, Eliza Dubleva from, uh, Coca Cola, Emily Xue from Scale AI, and of course, Marlon Duarte, who's heading hospitality and language services at ucla. UCLA Health. Um, and those are just, uh, some of the many panelists and speakers that you can expect. It's going to be a great show. Looking forward to it. Um, you know, it's not going to be 100 degrees Fahrenheit, I guess, near to see. So, uh, yeah, it's going to be. It's going to be great.

Speaker B: Well, there'll be a breeze.

Speaker A: At the very least, there will be a breeze. Please, please let there be a breeze. I need a breeze. Um, yeah, I have Two fans running at the office and uh, yeah, that makes it marginally livable. Cool. So today we want to talk to you about, uh, a few things. First, want to talk to you about, uh, the 50 under 50, our annual kind of startup monitor that we are putting out. Um, I want to talk a bit about at the other end of the spectrum, what some of the big technology companies launched recently. Just very, very briefly. There's these, uh, you know, always this flood of announcements that new language AI features are coming out. You're going to walk us through a couple of M&As and funding stories. Then, uh, there's some RFPs going on and hiring going on in public sector land. And uh, yeah, that's it for today. We want to keep it quite short because, yeah, as I said before, I think brains are probably at, I don't know, for me it's maybe 70% today. So I don't want to bother you guys with, with half takes. Literal hot takes. Actual hot takes.

Speaker B: You're right.

Speaker A: All right, uh, so the good thing is that for the language 50 under 50, we're not relying on hot takes, but on coolly, coolly, uh, elaborated research by our research team. So again, what is it? It's basically a collection of startups that are younger than 50 months. And we again found a cohort of 50 companies in the language AI space. So very, very happy about this. Can you remind me of the categories? I think we have 1, 2, 3. Hang on, seven categories?

Speaker B: Yeah, we do, yeah. Uh, I mean they change, they shift the categories over the years so they're not static categories as you might expect. Obviously. I think this is the third, fourth edition of the Slater language AI 50 under 50 that we've done now. Um, and yeah, it's very much based on the companies included, uh, and then looking at what kind of categories, uh, they naturally form. Uh, so we have translation and localization. We have multilingual media as a category. We have live communication as a category. Transcription and Captions, then voice AI and AI sign and data and evaluation. Those are the seven categories on this year's language AI 50 under 50.

Speaker A: Seven categories, seven takeaways. So let's go through the high level takeaways here and then explore them in a little more detail. So first, uh, one of the takeaways is that language AI startups are solving full buyer problems, not language tasks. And we're going to elaborate that in a second. Uh, the second take that our research team found was agentic and orchestration. Native startups are arriving now from day one, um, more specialized AI first LSIs are beginning to emerge slowly. Then what you mentioned before, AI sign. So AI Sign Language is becoming a genuine startup category. It used to be super, super niche which is a handful of companies and now it's a genuine category. The learning loop, the data learning loop is increasingly becoming the moat. Um, there is still room for niche specialists which is what's needed on top of these super powerful uh, general models. And then capability still matters, especially in voice. And we're going to talk about this in a second. Let's go to the first takeaway here. Language AI startups uh, are solving full buyer problems. Not language tasks necessarily. In the early language AI wave, uh, it was focused on translating text, converting text into speech, producing dubbed speech. Year's cohort, our research team found that language has really packaged inside solutions designed about much broader business objectives. Right, so let me give you uh, three examples. So the first one is a company called Suno AI. Suno with an H at the end. So uh, they offer a clinical workflow platform built on AI transcription, so speech to text. Uh, so it's generating medical documentation, assists with order entry, uh, electronic health records, etc. So suno AI. Another one is uh, verbalate LTP, uh, called verbalate. They're not really, I mean you'd think they're selling AI dubbing, but what they're positioning themselves is selling global reach. Right. So they help you and educators and creators transform a video into multilingual content. They basically give you the kind of all in solution that you can uh, make sure that your views go up and then Mix Halo. I think that's how you probably pronounce it, right? Mix Halo is Mix Halo or Mixalo, I don't know, probably Mix Halo Halo. So they have low latency audio delivery, real time translation captions, generated summaries et cetera. And uh, it's basically, it's saying they're a communication platform and DBL actually uh, acquired them, maybe acquired them, acquired them in June 2026. So. Well I guess that's the.

Speaker B: They just announced, yeah they just announced like a week ago.

Speaker A: Um, and I think they're based in San Francisco. So Mix Halo, if you're listening, you know where to go in September. Um, cool. Don't want to spend too much time on the agentic orchestration native startups complex. 1. Go to the um, the um, the 50 under 50 page and read about it. One, one thing that I do want to highlight, uh, there is a company called Hypro Localize which was apparently founded by the creator of Canvas Magic Translate. So somebody from Canva said hey, I'm going to do this, I'm going to start my own company and sort uh, of hyper localize with an S. And they focus on automating the work surrounding translation, context retrieval, brand consistency, UI constraints, review decisions. So anything around probably the design element. Right. Um, more specialized AI first lsis. So there's a company called uh, are beginning to emerge. So These firms combine AI native delivery with expert oversight. So the classic LSI. Right. Uh, so there's a company called Global3. Uh, they reflect that trend in media. It was founded by not uh, by somebody from Canva, but somebody from uh, Deluxe Media. So check out Global 3.

Speaker B: Still a lot of credibility, still a

Speaker A: lot of credibility in that space. And then uh, one of the takes number four, we had AI sign languages becoming a genuine category. And there we had a company called Rylo with a y. They raised 4,44 uh, 85 million uh dollar in funding recently uh, for, for that type of tech. And you know, unlike many language AI startups that build on top of uh, foundation models, I think in, in sign language companies often develop bespokes models from the ground up. At least that's what our research team uh, found. Did you know some of these companies? I mean I didn't, I didn't have Rylo on my radar.

Speaker B: Not many to be honest. Um, yeah.

Speaker A: What were the other coda. Sign. Sorry, Elephant. Elephant Talk and talk Sign.

Speaker B: Talk Sign I think I'd heard of. Yeah.

Speaker A: Cool. All right guys, welcome to the 50 under 50. And then another one is more on the data. Uh, it's uh, the learning loop is becoming the moat. Uh, so there was um, Basically the learning loop is becoming a form of proprietary IP built not on public benchmarks but on an organization's own data, decisions and expertise. There we have companies like Lenico, um, that's an AI translation platform that learns from consumer corrections terminology, domain specific language. Then a company called ownvia, they turn corporate terminology style guides, linguistic requirements into reusable AI workflows and agents. I like that. Cool. And another one is there's still room for niche specialists. So let's go through the room that the niche specialists. First one is lipped up. So single technical challenge, they focus lip synchronization. That's pretty niche and probably not soft yet I'm sure for kind of frontal, like. Yeah. Kind of YouTube videos with like people like us, like basically on a podcast. But there's many more complex scenarios in

Speaker B: like in the works for so long. I remember researching into it sort of five plus years ago. The whole like lip sync and Google doing a lot in that. I think at the time we had

Speaker A: lots of the um, blanket on the name. But we uh, you know we met some people last year in San Francisco. There's um, uh some companies in this space but yeah, it's complex. All right, let's go to the last one. Capability still matters, especially in voice. That was our take. So um, capability speech in voice AI is perhaps the clearest example for that trend because unlike in text translation where high quality multi, multilingual capability is increasingly available through foundation models in speech systems, it remains sensitive our research team says to latency personalization, multilingual alignment, voice quality, real time performance. So you have specialized model um, developers there that really um, continue to pop up. So we had in this year's cohort Cartesia, uh, and then our friends from New Phonic who uh, we met in London a few months ago and then Gradium who joined our panel in uh, in London and a company called Hamza is also on our list or Maya Research NC Speech and Shunya Labs. And Shunya Labs focuses on, you know, they focus on speech models and infrastructure for multilingual and regional markets. So there you have it, uh, seven takeaways from seven categories from our 50 under 50 list. Um, it's for free, you don't need to pay for it. You can go there, you can look at it, you can filter, you can enjoy it. Uh, and uh, yeah thanks to all the companies that participated and uh, hope to meet many of you, hopefully have some of you on the podcast. Um, yeah, great stuff. So do we want to talk about what is it? I uh, think we had it on our agenda but like so the Google expanding the AI Live speech translation again. Okay. We've, we've talked about it so many times. Uh, I think this is following on the news that OpenAI launched that real time uh ChatGPT, GPT Translate and Zoom launched like a translator and a summarizer. Now Google is launching AI live speech translation um in through Gemini 3.5 live translate. So tweaking the model, I think these big companies are taking any opportunity to do another launch. Right. So you go from 3.0 to 3.5, you do another big launch and people like us talk about it.

Speaker B: Yeah, I mean I suppose it's kind of keeping, yeah, keeping regular attention and focus on them, isn't it? As opposed to the competition or as

Speaker A: well as the competition and I m mean just to point. You did an article. We also followed uh the uh, worldwide developer conference, Apple's worldwide developer conference, uh 2026. Uh and we combined or kind of uh. Yeah, well we compiled all of the language AI translation, localization, language access capabilities that um, they announced. So there's some stuff for developers in there. And I uh, think we said that Apple demonstrated more advanced agentic localization workflow. Apparently there's something called uh, uh they showcase an example where Xcode translated an app into um, into Filipino, configured localization resources, updated strings and ran translation and accessibility tasks in parallel. So yeah, agentic automated. Finally um, we unpacked the Mythos. How would you pronounce that, Claud? Mythos. I'm having trouble reading this. It's one of those words that like I look at it and it's the same in German and I just can't make and so I'm reading as mutos. But it's Mythos. All right, so there's this whole story around like uh, Anthropic launching this model they've been warning about for like half a year to kind of you know, do the fear pr and then they launched it and then the US government told them to pull it again for whatever reason. There's probably a million other podcasts that explain that to you. But what we did, we went to the so called System Card and looked at uh, the language uh AI capabilities of this amazing too dangerous to release or too dangerous to keep online rather uh, um, model. And uh, yeah there's some uh, fairly interesting takes there. Um so you know they um, the System Card did include some notable language related findings. They benchmarked it in um, in 44 languages. It performed really strongly across a number of like very geeky kind of evaluation benchmarks that you know, if you're interested in, go to the website and look at it. Um, and yeah there's, I think we pulled out one quote by uh. Who's that from? Did, did you pull that in into the notes there?

Speaker B: There was a comment by RWS, uh head of innovation of train AI by RWS, um publishing on uh a post on LinkedIn, just talking about I suppose the, the release. Um and he Tomash uh characterized it as more of a catch up than a breakthrough in multilingual AI. So he said that Fable 5, it does reach parity with leading frontier models but it's not kind of obviously or clearly surpassing them.

Speaker A: And I don't understand what's the mythos 5? Fable 5. All right, again there's other podcasts. So big model Anthropic Fear PR for half a year now. Major story about the US government pulling it. And if you rely on um, the expertise of the head of innovation at ah, train AI from rws, it's basically catching up, not exceeding at least in the language component there. All right, do we want to talk about um, investments? A couple.

Speaker B: Yeah, we definitely should. I mean some of it will be familiar from what uh, actually what we've covered either in the 50 under 50 or um. Well in particular, um, so I mean you mentioned, I think Rylo, um, we covered on the one hand it went into our uh, 50 under 50 but um, as you also mentioned it raised the company raised 85 million um, recently. So we had a coverage of that um, on Slater. So they've got 85 million to build AI language solutions for the deaf population. Um, yeah, a lot of money. Um, and as we said, as we said in our analysis, this is becoming sort of a very credible category now or a bonafide category. Um, then we also covered Dell Technologies Capital. Yeah, exactly. Um, Dell Technology Capital, uh, uh, lead an investment into bland, uh US$50 million into bland, which is a full stack voice AI startup. Uh, as we mentioned, Dell Technologies Capital will be on our investor panel at slatecon San Francisco in just a few months. So yeah, hopefully you'll get a chance to kind of unpack a little bit about that live on stage with them. Um, but what can I say about uh, Bland, uh, the. Well based in San Francisco. Great. Why don't they come along as well? Um, the startup builds voice AI agents for complex enterprise phone conversations. This, I mean I've seen this multiple times, numerous times when we're talking about sort of um, uh, areas or startup areas that are getting investment currently. Um, they're sort of. Well from what I can see, I mean their shtick seems to be taking or handling longer, more complex phone calls than maybe many of their peers. So they say that their conversations that they're handling often last um, between 30 and 45 minutes. Obviously that's a long time to be um, for the voice AI agents.

Speaker A: Even the AI gets tired.

Speaker B: The AI gets tired, yeah. Needs to think for a bit longer, um, and then um, well, smaller scale in terms of the investment. Um, we had a news story that Gridly, um, which specializes in kind of gaming and content, uh, management, uh raised 1.5 million um, for their content management platform. Um, with the idea being they're going to be using that to build agent, agentic AI into the platform. So again they've got this specific. Well they have a, they have a clear niche which is the gaming localization. But I think they say that they're also um, likely to start targeting other sectors or kind of bringing the technology to other sectors such as FinTech, EdTech and B2B SAS. Um and I think you spoke to their CEO actually and uh, Albinson on the podcast.

Speaker A: Uh, yeah, about a year ago. A bit more than a year ago. Yeah.

Speaker B: Yeah. So those were sort of the three um, funding stories I wanted to highlight. Yeah, so we do have a couple of M and A stories to bring to you as well. Um, we, I think both, both happening within Europe. Um, these two since, since we last spoke. Um, so first up is Powerling, um, which announced it's acquired Atlantic Traduction, uh, which uh, is a France based LSI that's been operating for more than 45 years. Um, and they have experience in defense, naval industrial and technical sectors. Um, seems to be fairly sort of boutique. They had uh, when we looked something like nine people on LinkedIn. Atlantic translation. There you go. I localized the company name when I shouldn't have done it.

Speaker A: There you go.

Speaker B: So that was a, that was um, an acquisition by them. I think it's probably their first, the first acquisition that they've completed since uh, they acquired Edem um in California back in, well last year that was um, which we represented in its sale. Um, and the second piece of news uh was that Digital Talk in Sweden has acquired Hieronymous uh in Switzerland, Zurich based legal translation firm, ah, which was also represented by Slater. Um, so it's been a. Yeah, it's been a good time. Um, again so this is Digital Talk second acquisition after buying 24 Translate UM in July 2025. Uh, also repped by Slater, um, uh, and yeah, legal specialist, um, now part

Speaker A: of ah, Digital Talk building a Switzerland presence. All right, so uh, you have a few more weeks to register at the early bird for Slatecon San Francisco. Um, you know maybe like three or four weeks. Um, so catch that early bird while it's still there. Again, amazing agenda in the build up and we'll leave it at a cool 20 minutes here to uh, make sure that we survive the heat wave. Apparently neither west or yet. The camera here on the iPhone is running hot. Cool. All right, thanks for listening and uh, we'll be back in a week or so.