This is an automatic audio transcript made from the following Slator interview. We used our new feature “Vocabulary” to spell out the speaker names and the company name Weglot. This transcript is not edited by human. 100% automatic transcription. Review and make a decision!
Augustin (00 : 03)
We’re building something that’s used by 60,000 websites over the world.
Florian (00 : 09)
Press releases are being super lightly post edited machine translation.
Esther (00 : 14)
Now, a lot of the translation has actually been copied from a fan made translation which is known as gambling.
Florian (00 : 30)
And welcome, everyone, to Slaterpod. Hi there, Esther.
Esther (00 : 33)
Florian (00 : 34)
Today we’re speaking to August Poor, co founder and CEO of Weglot, a fast growing web localization technology company based in Paris, France. Very good discussion. Very interesting discussion. Learned a lot about web locks. So do stay tuned. Esther, today is an exciting day for us. We’re launching our 2022 Market report. We briefly test it last time, and today’s the day.
Esther (00 : 58)
Florian (00 : 59)
We do record this on a Thursday. So by the time you’re listening to this, it should be on our website. Yay. But before we go there, let’s go through a few kind of AI machine translation bullet points, and then we go over and talk to August. So Google’s huge new language model answers questions, does other stuff. And we’ll try to give you, I don’t know, the key bullet points there, although it’s a huge paper and it’s a huge launch. Then you’re going to talk about scandalation.
Esther (01 : 32)
Florian (01 : 33)
And issues in the animated translation world.
Esther (01 : 36)
Florian (01 : 37)
And then we’re going to close on a uni buying another company. And then there’s a surprising post editing machine translation plot twist for you listeners. All right. So, hey, this week in AI news is that AI draws and AI writes and AI answers questions and obviously helps us with translations and all of that. But let’s dwell on the drawing point. Did you see all those weird AI drawings by some new model that just came out this week?
Esther (02 : 07)
I didn’t, but I hadn’t. And now I have. They look very interesting and colorful.
Florian (02 : 14)
They’re kind of weird scary. I forgot the name. And we’re not going to talk about that. But it’s just basically on Twitter. All of a sudden it blew up, like in the last few days about just breakthroughs in AI. And of course, one of them was language, and we’re going to talk about that in a second. But the other one was another model that’s doing that draws right or whatever the appropriate term is. So you can say, like paint. The one I remember from this morning was like, what was it? Rabbit on a bench in Victorian Times, reading a newspaper or something. And then the model grew that rabbit on the bench, Victorian style, reading a newspaper. But there was all these weird things on it. So check it out.
Esther (02 : 56)
I wonder if all of the word illustrationists illustrators are getting nervous that they’re going to be replaced by machines or whether there’s some kind of post editing style workflow to be found in Illustration 100%.
Florian (03 : 14)
Super interesting point. Go on Twitter. The exact same discussion as in language. It was literally there was the predictable all the illustration guys are going to be out of work. And then there was the other guy was like, no, it’s a tool. It’s a tool for them. Right? So there was this exact same dynamics. We’ve had this debate. We’re there. That’s why I keep saying at these presentations we’re doing that basically, we’re very much ahead of the curve when it comes to humans working together with AI. Because for the illustration people, this is breaking right now. Cool. So on, the AI writes and answers questions and translate side. Well, it’s another one of those big language models. This time it’s got 540,000,000,000 parameters for breakthrough performance. That’s what the Google blog post says. Now, can I assess if it’s breakthrough performance in translation? Absolutely. I cannot. But it does do a lot of things. This new $540,000,000,000 parameter models, and one of them is translation. And if you go to their page, to the blog post, it’s like a tree that grows. It’s hard to describe in a podcast, but it’s a tree that grows and it has all these use cases around it. And at 540,000,000,000 parameters, it’s doing things like dialogue, pattern recognition, common sense reasoning, logical interference chains, question and answering, semantic parsing, arithmetic, co completion, language understanding. I could go on. And of course, translation translation is actually quite a sizable thing there. So this new language model by Google does a lot of stuff. I wonder how common sense the common sense reasoning is there. But we’re heading towards that bigger AI. We don’t have to dive too much into detail, basically. Again, it’s kind of a cheap style. It’s just a Google version of cheapity three, if I understand it correctly, that does all kinds of AI tasks and translation is one of them. They parse it in a particular chapter of that 8090 page paper they published and kind of give some Blue scores and have a few observations. Like, the results have been particularly impressive in translating into English, but while translating out of English, it yields more lackluster results. We’ve had this discussion before around these bigger models doing translation tasks and people. We were told that it would probably never be as good as a dedicated model. But it’s interesting that these big tech companies continue to release these big models and probably something we need to be aware of. So before I dig myself further into this whole of ignorance, we should move on to something. But also briefly that I find super useful. Recently, I’m following a lot of the news that’s coming out of China, my Chinese is really not fluent enough to read Chinese posts and things like that. And so I use Google Lens a lot. Oh yeah, yeah. When you go also for what’s coming out of Ukraine and Russian, obviously I can’t read any of this. You can actually use Google Lens, and even if it’s an image, you use Google Lens to kind of OCR and then Google Translate to translate it. And it’s quite useful for kind of informational purposes. So Google Lens, something that was I think, launched like three or four years ago, I remember, but now it’s coming in quite handy, all right, away from Google AI OCR and big language models to the world of manga and animated translation. Esther, what happened there? Big scandal scooped up by Katrina.
Esther (07 : 14)
Yeah, well, it seems like a big scandal based on Scanlation. I tried a bit of word play that, but as you said, one of our previous Slater Pod guests, Katrina Leonidakis, she seems to be sort of central to the analysis of this and kind of was tweeting about it and was quoted in some of the coverage. So the issue seems to be that there’s this manga called Ranking of Kings and the English translation English release of Ranking of Kings has been temporarily suspended because of typos and translation issues. So this is a manga by Susuki Toka. It’s been sort of published in series of volumes I think for several years now, but now it’s also being serialized in a comic, Beam magazine and published in twelve different volumes. So the English translation is being done or has been done, and it had been published in seven different volumes as the kind of official version, and it was actually being sold in English since for about a month or two now. But apparently all of these issues have been found, which now means that these seven volumes, as a minimum, are having to be retranslated. People who’ve bought these, the seven volumes of Ranking of Kings, they can still read it, so they can still have access to it, but they’ll also have access to the updated translation. So once the retranslation has been done, I don’t know how big seven volumes of this manga looks like, but it sounds like quite a lot of content anyway to have to redo, not to mention the sort of like public embarrassment of kind of having to admit to some of these issues being in an official release English translation.
Florian (09 : 22)
What’s the issue?
Esther (09 : 23)
Yeah. So the main issue here is that a lot of the translation has actually been copied from a fan made translation, which is known as Scanlation. So I think often it happens in game localization. It happens in anime. Manga fans who are sort of like really hardcore into certain manga anime will provide their own versions, make it accessible to themselves and to the community. But now, obviously, the official English translation of the release has been commissioned, and it seems like whoever worked on the official English has copied quite indiscriminately from the Scanlan version. The article that we were looking at says it’s a bit of a legal Gray area because actually fan translations, these uncommissioned translations, if you like, are themselves a form of piracy. The team that did the original uncommissioned Scanlan version didn’t work on the official translations at all. So sort of plagiarism, I guess. So Katrina, who we had on Slate Spot a few months ago. Now, who is localization expert, Japanese to English, like deep expertise in anime, manga. She tweeted about this, saying that she spent a few hours analyzing the differences between the official release of Ranking of Kings and the scannalation. Obviously that came first. And she said 42% of all dialogue in chapters one to three of the official translation has been directly lifted from the Scanlan. So that was her assessment, as well as some of this copying plagiarism. There were also, I think, incorrectly used phrases and stakes, things like that. The English distributor and the translation provider have both apologized for a lack of quality and said that these issues might have caused serious damage to the quality of the original work. So they seem to be putting things in place to try and rectify the issues. But obviously a bit embarrassing if it’s already been sort of sold and published, being distributed for a couple of months now.
Florian (11 : 51)
That doesn’t happen in many areas where you have fan translation. No one’s going to fan translate a financial report.
Esther (11 : 58)
I was going to say an annual report, like just for all of these interested investors. Let me do your favor.
Florian (12 : 08)
Yeah, that doesn’t have been anywhere else. Interesting. And I love how this community is so active on Twitter. And that’s why how we came across Katrina in the first place, because this is like almost like public conversations taking care, taking place on Twitter, where they have like 2300 retweets sometimes on a very kind of seemingly niche issue from an outsider’s perspective.
Esther (12 : 29)
Yeah, there’s a lot of passion. I think there’s a lot of passion and feeling behind this.
Florian (12 : 33)
I wish we’d get 2300 retweets per.
Esther (12 : 35)
Tweet, but we don’t.
Florian (12 : 36)
So anyway, follow us on Twitter at Slavery News now, our friends at Auno, SDI made an acquisition, not in the lock space specifically, but tell us more. Yeah.
Esther (12 : 49)
So technology investment really in a nutshell. But SDI said they had acquired a technology provider called Autonomous Media Groups based in the UK. It’s kind of workflow management, they say scalable workflow management, asset management specifically for the media content side of things. Autona helps automate processes and media workflows and they say reduce operational costs. So, yes, it’s been acquired by SDI. The idea being to integrate autonomous platform. So they’ve got SaaS and managed service solutions, including I think one possibly their flagship one called Cubics. But all of that is going to be integrated with SDI to form end to end supply chain for media and media localization services. So it’s quite a smallish acquisition in the sense that seems like 15 to 20 people on LinkedIn. They are sort of selling around the world. They’ve got resellers in Australia, Europe, New Zealand, South Asia, North America and South America. So they’ve obviously expanded and done quite well. But in terms of subsize, relatively small company founder James Gibson, also the CEO, is staying on so autonomous going to be run as a wholly independent subsidy of Ian STI. So James is going to remain CEO, and he will also become the VP of Product and Architecture for Iunosdi reporting into. Iu Chief Information Officer Alan Denbri. So, yeah, interesting sort of tech focused acquisition there for SDI.
Florian (14 : 40)
If I was a German speaking media localization industry participant sitting in Berlin and I wanted to learn about this acquisition, I would be alerted by PR Newswire that this happened from a Press release in German published by a uni SDI. And I’d read it and I’d read something that was post edited using D Bell. So why do I know? Because when we read that article, there’s like an option in there’s, like a.
Esther (15 : 17)
Drop down, isn’t there? At the top of the prayer?
Florian (15 : 21)
Yeah, yeah, there’s a drop down. I went to the German version, I compared the source and then Google Translate and Dbell with the actual published content. And the first sentence is literally word for word, Mt. So not even a sprinkle of post edit then the second very long sentence, Sir, I’m just talking about one select paragraph because obviously I didn’t look at the whole piece, but one paragraph or one sentence of one paragraph was translated as one sentence as well by Google Transit. Almost identically, by the way, by Google Translate. It’s interesting how identical the two MT’s are. Now the actual published version, though there is a post edit component to it because the detail version was just super long. Like it was a very long, barely readable sentence. I mean, grammatically correct, but just like super long. So the post editor said a period and then broke the sentences into two. But really interesting that press releases are being like super lightly post edited machine translation now. Right? I think that’s remarkable because it is a Press release.
Esther (16 : 48)
But who’s paying for that then, Florian? Do you think? Is it sort of integrated with PR Newswire or is it the client SDI who’d get charged for that? Or is it all kind of bundled into the price of publishing a PR?
Florian (17 : 02)
I would assume it’s bundled. Pr Newswire was actually a client of mine. It’s like ten years ago. So I guess I can some stage in your life. Yeah, the previous LSD, I used to work for quite competitive rates, and I’m pretty sure it’s part of the bundle. And maybe you can order, like, which languages you want it to be published, but it’s probably part of a broader press release bundle if you’re a large company like SDI is just interesting from a text type point of view that press releases are now part of a category that gets the super light post editing treatment. I find this remarkable because you’re reading through the text and it is correct. I mean, the Mt is also, in a sense, correct, but it’s just like, as a German native speaker, the English just screams at you below the surface of the German like, just the way it’s phrased is super jargon heavy German, like stuff like highly scalable end to end localization supply chain. Yes. You can convert this into German words, but what does it really mean?
Esther (18 : 17)
I think it’s interesting from the perspective of thinking about publishable content. And what is publishable content, because press releases are published online and then you can reference them via URLs for years to come. And in fact, sometimes we would quote from press releases when we’re sort of digging into the context behind certain things. So they do have a shelf life. They don’t vanish entirely, but I suppose they become less relevant after that kind of initial news hits.
Florian (18 : 53)
Absolutely. Also, then you start scraping the web for parallel content and you’re basically scraping super lightly post edited content. So it’s this kind of machine. And then the AI learns off of it and it’s giving from the post edited content from the light post edit. The post editing was so light, it’s literally like one like it’s just breaking up one sentence and then making it so it’s grammatically still correct after you broke up that sentence. Right. That’s it literally. That’s it. There’s almost zero. I mean, the edit distance was super small here.
Esther (19 : 28)
And I suppose what you’re saying is if the press release had been drafted in German, it would read quite differently.
Florian (19 : 35)
Yes, I think so.
Esther (19 : 38)
It’s kind of synthetic in the sense of if you’re using it for training Mt purposes, you wouldn’t want that to be okay. This is German source content because it’s not really an accurate reflection of German writing for press releases.
Florian (19 : 53)
100%. Yes. I mean, there’s a few of these German long words that the Mt has created that there’s no way you would even come up with that word in the first place. It’s not technically like a translation error, but it’s just like it’s such a long word, like you read and you get it, but you’re like, yeah, it’s not a word. That would be my active vocab. Right. Anyway, on that good observation, we shall head over to Augusta and talk about web localization.
Esther (20 : 23)
Florian (20 : 31)
And welcome back, everyone, to Slaterpot. We’re really happy to have Augustine Paul here. Join us. Augustine is the co founder and CEO of Weglot, a no code website localization tech provider. And they got people’s attention by raising a cool 45 million Euro from gross investors recently.
Augustin (20 : 47)
Florian (20 : 47)
Hi, Augusta. Welcome to happy to have you on the part.
Augustin (20 : 50)
Hi, Felon. I’m really super happy to be there, too.
Florian (20 : 54)
Awesome. Great. Where do you join us from today? What city, what country?
Augustin (20 : 59)
I’m joining you from Jarrett in France. The company is based in Paris, but I live in Paris and I go back and forth to Paris.
Esther (21 : 07)
That’s nice. Part of the world.
Florian (21 : 11)
Some good surfing there. We just reminisce before we got online here that I used to spend some time there in the summer. Amazing place. So, August, tell us a bit more about your background. Like you were with an investment bank. Lazar. Right. And so how did you transition from the investment banking world to web localization? That must have been quite a twist in the turn.
Augustin (21 : 36)
Yeah, exactly. Yes. When I was in Bank, I didn’t know anything about translations or web, actually. So I spent three years doing major acquisitions, and I really enjoyed it. Super intense environment. But at some point, I started to get bored and I started to naturally want to go to the office in the morning. So I thought, okay, it has to change. So I wanted to find a new challenge. And I thought starting a company or joining a company super soon could be the right path for me. And at this time, I started having a couple of ideas in my head and trying to test them and also meeting a lot of different people who had ideas at this time. And that’s when I met Remy Wiggle, co founder and CTO, who had the idea of a first user and the first MVP. So when I met him, I didn’t know anything about HTML CSS whatsoever, and not anything also on translations, ASP or anything. But when we had the first conversation together, he explained to me how he had the idea, the challenges he faced as a developer. And that’s how I really got into this Wiggle adventure.
Florian (22 : 59)
That’s like a business cofounder technical cofounder combo, right?
Augustin (23 : 05)
Yeah, exactly. Remy has an engineering background. He did consulting for finance for a couple of years, and then he worked in web company, like real time billing, like Critio, but the US competitor of AppNexus. And then he quit and he started the first company, which was kind of a classified app with Google Maps, so you could see on your app what people were selling around you or buying around you. And he did that for a year with a friend and cofounder. And after a year, it was super hard to raise money. It was a free premium model, super high competitor in France. So they decided to just shut down the company. But when he shut down the company, he thought about the different challenges he had when he actually did his first entrepreneur, Johnny. And every time he met a technical challenge, he had a super easy solution provided by a third party that was only doing that. For example, when you want to add payment to a web application, it’s not easy. Will you do the connection with the bank yourself, hosting the bank account and so on? No, you just use Tripe. It works out of the box. It takes like a day to integrate. And that’s kind of magic. And he found the same thing for search with algorithm or for text messages with truly and so on. So, yeah, every time he had and he met a technical challenge, he had this magical solution but when he had to do the translation, the web app, he didn’t find that magic. And he actually had to do a lot of the technical work, super time consuming by himself, which is rewriting the code, making sure that are working, having the button, making sure certain Giants will index, see and rank the page, and so on. And it really took him a lot of time. And really the pain was coming from the technical part on. The content part was quite simple, strings and sentences. So not that hard. He spent a couple of years in the US, so he knows how to speak English better than me, probably. So, yeah, it was coming from a technical pain point. And he thought there should be a magical solution just to help any web developers, website owner to make a website mission and gold in minutes. That’s how he presented me the idea and what he was working on. And I was sold from day one. And so I don’t know how to code. How can I help you? I’m going to find users and we’re going to see if it’s working and people like it.
Esther (26 : 13)
Yeah. So I was interested about that part. Really. So you said that’s obviously the backstory to Remy’s idea or concept behind Weglot. But what was it about the way that he pitched it or the opportunity that really appealed to you? And then also tell us since then what’s your journey been like, any major pivots or some of the key milestones you faced together?
Augustin (26 : 39)
We did not really pivot to be very transparent. And actually the vision he had and presented to me at this time is the same today. It’s really about making this translation feature through this solution. So our goal is that we got is the translation feature for websites, translation. And that’s really how we see things today. And that’s how we saw things back then. But obviously it was not super linear and easy. So the first hard thing was to find users. So how do we find people just using the product and understanding what they like, what they don’t like? And we quickly found out that two things were super important. One is it has to be super easy to integrate. So at this time, there were no local, no code trends. It was really about just okay, I have a website. I’m not a technical engineer or a developer. How can I add your product to my website? So that was one thing super important and the other one was okay, it’s working. But will search engines see the translated versions? So you can just do translations in the browser on the fly. Otherwise search engines will not see it. So you will clearly cut yourself from the big benefits of having a website process it. So that’s what the two things. And it drove us to being first and finding traction inside the WordPress universe that you might know at the content publisher, you might choose WordPress.
Florian (28 : 24)
We’re on WordPress.
Augustin (28 : 25)
Okay, so you’re on WordPress. So we found our first traction in WordPress and it worked really well. Then we did that also in other CMS, which is Shopify. So it’s more online stores, ecommerce. And then we finally found out a solution that we were able to add to any website, regardless what technology they’re using. So today, if you’re using Shopify, Webflow, WordPress or any other CFS you can use with that super easy. And if you’re even using a custom solution, it’s also possible.
Florian (28 : 58)
Let’s talk about the funding you raised from Fun called Parttech Gross. I think as we wrote, it was a 45 million round. So tell us a bit more about that. What was the decision making process behind accelerating what you’ve already been doing through raising funds? And maybe was there a prior round you had or was it bootstrapped up to them? Just gives a bit more kind of color on that.
Augustin (29 : 21)
Sure. So we started with that in 2016 and we did a small precede or sit around of €450,000 in 2017 and since then we didn’t do any raising. And so we thought maybe it was time to partner with Zip new guys like Patek. The goal was first to two or three fold. One is let’s find people who know how to support companies like us going from our stage of growth, which is like 10 million error to the next 1000, very tech oriented global positioning. And that’s what they do. They really do this as business with SMBs, with the type of companies, with our level of growth. And the other one is it was important for us to be able to take more risk now, not to transform ourselves into cash burner because I think we don’t know how to do that, but maybe being more concurrent. So we have more resources to take the market and to penetrate even more the different markets we are already addressing and addressing new ones. And the last one is also hiring people. It’s really about having strong Empire brands and having great talent to build new know how we want to develop.
Florian (31 : 02)
Just you mentioned people there. So Where’s the bulk of the staff located? Most in France. Are you fully remote or how’s the team set up?
Augustin (31 : 11)
Most in France? Only in France. We have eight nationalities, but we’re all in France based in Paris. Some people of the team are based in other cities like me, but most of the team is based in Paris.
Florian (31 : 27)
Cool. So let’s talk about the client segments. Right. So what type of clients have you attracted early on? Where’s kind of your core base right now? And are you planning to go more towards the enterprise side of things that are like very complex deployments or really kind of the SAS layer, more no code layer. Now just talk a bit through those client segments.
Augustin (31 : 52)
Sure. We’re coming from a very self service, small SMEs that we love market and it works very well. And we were only doing that until beginning of 2020. And beginning of 2020 we started to see more larger companies coming to us with larger needs or they wanted to have someone to help them to understand the value of the product before they can eventually do a park. And that’s when we started the enterprise segment. And that’s really about providing the same product with more usage or more needs and more service. But it’s the same product. We really want to have this idea of we’re offering a product, not a service. We are not LSP. We are really a solution that helps you to make your website translated. But we’re more partnering with RSPs. I mean many of our customers are using professional translators with that. And the goal is to keep doing that and to grow the two segments. The self service part SMBs, but also the enterprise one previously going more and more into the enterprise. I mean, one thing they love is that the more they have technical depth, the easiest it’s for us to be used because we’re kind of a layer or any frustrator that you would plug in on top of what you have and it’s working out of the box.
Esther (33 : 24)
This kind of low no code movement thing, for those who aren’t really super familiar, tell us when did it all kick off? What are some of the drivers has coveted had any kind of impact on the low no code movement? And how does it all fit into the web localization universe?
Augustin (33 : 44)
Yeah, it’s interesting. So when we started with lots like I said, there were no code, no code words at this time. But what we had in mind is really about we need to minimize the time between the discovery of relat and the while we affect and the value we provide you. So we need to be very good at when you start your signing process. You need to see transaction versions of your website super fast. So removing friction with technical back and forth or anything like this was super important to us and it’s still super important to us. So that’s one thing and that’s what you can call no code, which is for us more. We are building complex things and taking the complexity for us. So it’s super simple for our users that’s I think what is the local no code? And obviously it’s really highlighted and accelerated with Covet and with Digitalization of course, and more and more nontechnical people are in charge and responsible of web applications, websites and so on. And that’s also another reason why such tools like us are relevant and more and more used. And the other thing is that we are actually at the crossroads of two pinpoints. One is super technical. So tomorrow if I ask you like put your website in Spanish and Chinese. Okay, so there is a technical part that’s complex and the other one is contempt. Okay, I don’t speak Spanish and I don’t speak Chinese, so the maintain is huge. So if we come to you with a solution that’s helping you to do 80% of the work in a couple of minutes, it’s huge value. And that’s why it’s I think also one of the reasons of our success that it’s super easy to immediately have 80% of the job done. So you can focus on the 20% parts if you want.
Esther (35 : 40)
What about some of the kind of complexities of website localization? How do you deal with things like SEO you mentioned? Sometimes there’s an issue with or there can be an issue with Google not recognizing the translated version of the website. What are some of the main challenges around that?
Augustin (35 : 58)
Yes, it’s super important to us and that’s one of the first feedbacks we actually got when we started with that. So we are good students. We read the Google documentation to understand what was important. And actually, technically speaking, there are three things that you have in mind. One is having your transitions on the server side. So it means that it’s rendered by the server and it’s not only in your brother. For example, if you go as a visitor to a website and you see sometimes the brother is proposing you to switch the language and you can switch it from English to French, for example, but it’s only in a brother, so it’s not in the source code. So that’s one thing. The other one is having dedicated URLs. So you should have a dedicated URL to indicate Google that there are two versions of the page. For example, you can use subdomains mywork.com for English and Fr myworks.com for French. You can also use top level domains or February. And the last point, super technical. Sorry. The last point is help Google know that they are different versions of your website. And there are two ways to do that. First is to have a site map where it’s basically a map and it says there are different versions of the website. And the other one is to add edgereflong Tags. And both of them are the same purpose, which is let Google know that there is alternate versions in other languages of the page. Jewel in his crowd.
Florian (37 : 37)
Esther (37 : 39)
Yeah, I’m making notes as we go, listen and learn.
Augustin (37 : 43)
And we’re doing that for you out of the box again. The good is that it’s easy for you. You can just focus on eventually work iterate on your keywords or things like this. Not on the technical part, from the.
Florian (37 : 55)
Technical to the language part. So you guys don’t offer the service of translation, right? So you’re partnering with LSPs or your clients would bring and on board their own whatever freelancers or LSPs, is that correct?
Augustin (38 : 09)
Yes, exactly. I mean, our goal is to offer our users best resources to do their own translation workflow. Quality, depending on their resources, the time they want, etc. So what we do is that by default we offer machine translations so they don’t start from scratch, they can actuate the display or not, they can change it or not. Then they can edit that themselves with their local teams or own profitable localization team, or they can invite their LSPs or they’re working with to do the edit and the review. Or they can outsource part all of our translations to professional translators that we are working with today. We’re working with TextMaster. So Text Master is a marketplace owned by Icloud, but it’s also possible to do, export and bring in your own LSP if you want. The goal for us is really to be able to help you give you the resources so you can do the thing you want.
Florian (39 : 14)
The translators can work in Weglot itself or they won’t.
Augustin (39 : 19)
Today we don’t have a marketplace built inside Weglot. But what you can do, for example, you can invite to your translator for a specific language, you can even assign translation to them and they come in to the account, they can review it, they can see it on the web page of the context, just transitions, and you notify on the turn and it’s live already on the website.
Florian (39 : 44)
What’s your customers perception of machine translation like in 2022? Because there’s probably like a wide variety. People think, well, it’s basically a click and then it’s done and others maybe have a bit more of a nuanced understanding.
Augustin (39 : 58)
But yeah, I think it’s really varies. It depends on use cases. For example, if you are an online ecommerce store and you have hundreds of thousands of products, it’s not going to be possible to do manual human transitions. I mean, it’s just not scalable and it’s not a very big driven. So generally speaking, ecommerce tend to use machine transitions by default and then iterates on most profitable or seen or most important pages. And then you also have like, for example, another use case, it could be Coffee website with a marketing website that’s really about the coffee voice and turn, and it’s super important for them to have that in the different languages. And for them, machine transition can be a resource and tool, but they need to really validate and make sure it sticks to their constraint, which is good. Again, we’re not recommending anything ourselves. We’re just letting them build what they want. And now going back to the perception of machine transition, I mean, mine when I use Google translates when I was in College, it was horrible. It improved. I’m so impressed today with the quality it offers for some types of contents. It’s super impressive. It will never be humans for sure, but it’s really a great tool. Absolutely.
Esther (41 : 35)
I mean, thinking about sort of you mentioned you’ve got eight different nationalities, but all based in France at the moment. How has it been in the past couple of years trying to hire and retain engineering talent. Tech talent on the one hand, obviously it’s really competitive for talent at the moment, but also with Covert, I think makes life a bit more challenging as well.
Augustin (42 : 01)
I mean, it’s changing. I will not lie. But yeah, overall it went well. I think that also the mission and the opportunities is very interesting. We’re building something that’s used by 60,000 of websites around the world and we have a unique opportunity to create a brand which could be the feature for transactions for the web, which is I think quite exciting. We’re using state of the art cloud services, so it’s also attracting engineers to join us. Also, we’re kind of picky and we’re not so good at anticipating. We try to improve ourselves, but we tend to wait until we’re very under the water before starting a new job offer. It’s changing. We were 30 people, so that’s not huge team. So I think it’s less challenging than for other tech companies that are 400.
Esther (43 : 13)
People and hiring outside of France, potentially.
Augustin (43 : 17)
No, not yet. For now, since we’re a small team, we think it’s super important to share the culture and we are not remote by default and from the very beginning. So we don’t have really a culture that’s super easy to, I think to build and improve with the remote only environment. So for now, maybe it will change one day. But we hire in Paris, in France.
Florian (43 : 46)
So it looks like when you started out it was mostly technical roles like now with part tech onboard and kind of a more aggressive, I would assume marketing and sales strategy. Are you hiring more on that side of the business and generally what’s your marketing approach been and where do you see it going now? Because you seem to have great success onboarding clients now through SEO and just other channels. Right. But how is that going to change going forward or double down?
Augustin (44 : 15)
Yeah, we’re going to double down for sure.
Florian (44 : 20)
Augustin (44 : 20)
Different things first. We’re also still hiring in technical positions. That’s super important for us and also in support, which is a mix of technical and business on the marketing and sales part in sales. We’re also hiring technical people because sometimes it’s important. But yeah, double down what we’re doing. Also the exciting thing is that we’re discovering more and more usage over time. For example, we get interaction with, for example, local governments or government websites, which is I mean, they have a big challenges to be accessible and to be translation compliant with their own policies. And so that’s a new use case. So we need people to just be able to absorb the demand. So it’s really about absorbing the demand and also building the road to market and double down on what’s working. The new thing we want to build is probably being able to build the greater Brennan Wallace inside marketing communities, inside localization people communities, inside these types of communities that are less technical than the one we were used to talk to in the past.
Florian (45 : 41)
Got it about growth in general. So now you’re super firmly in this kind of web ecosystem with the WordPress you mentioned. And I think Shopify, are you planning to add or have you added other kind of web whatever for like a better word, other ecosystems or CMS like side core? And then beyond that, what could be growth errors or you’re happy with web generally?
Augustin (46 : 07)
Okay, so one day maybe we could do native mobile app, but for now, it’s a logic that’s a bit different. I mean, the way we’re doing translations for websites, it’s really real time synchronous and mobile native mobile apps are not in real time by nature. So it’s another play. So for now, we think that the market is super large. The web application and website market is very large. So we should just focus on keep improving the product. Really. We’re focusing on solving these paints and trying to offer the best solution for that. And as long as we have room to increase our market share and being more present, we will first focus on that. And then maybe one day we’ll do.
Esther (46 : 58)
Something else and you say it’s a large market. What’s your outlook in terms of growth and trends and drivers for web localization.
Augustin (47 : 07)
So for us again, we are the crossroad of translations, localizations and websites. So there are more than 1 billion domain name registered and it’s growing. And I think that the web translation, the online and web pages in the translation industry is growing. So there are more and more demand for these types of format. So yeah, we’re on two super great current, but in the right direction. And yeah, I don’t have a specific number. I could say like, okay, it’s a 15 billion USD market probably, but I think it’s a large market that is growing, which is exciting to close.
Florian (48 : 05)
Tell us the top two to three things that are on your roadmap for the next 1218 months, features, additions, new things, anything you can disclose or want to keep it under wraps.
Augustin (48 : 17)
I mean, I can already discuss things that are either in beta or about to be launched. First is we’re having a new integration with Square Space that help Squarespace users to easily use us inside the admin products of Squarespace. So they can just activate with us in that inside that. The other one is we just released a super exciting feature for us. I don’t know if you will share this excitement, but we can now translate variables inside. We let it mean that customer X buys N product. Now it’s just one string and it’s not N strings, for example. And the last one is we really want to be this translation infrastructure. So it’s important for us to be able to offer the maximum of flexibility to our users. And it means in terms of URLs, they can play with URL so for example, they can have subdirectory that could be Fr but if they want they could be Fr B e for Belgium so they can have local original versions of their language. They want and that’s something that we’re working on and that will hopefully be ready this year.
Florian (49 : 37)
I got a square space site I can play around with. I’ll check it out I’ll check it out when it appears there. Cool. All right. August you so much for doing this. This was really interesting and good luck to you with all that the new partnership with Partech and your plans. Thank you so much.
Augustin (49 : 53)
Thank you very much, guys. Happy to be with you.
(49 : 55)
Transcribed by Gglot.com