Build a multilingual transcription bot with ElevenLabs #Scribe API

Transcript

We just launched the world's most accurate speech recognition model, achieving industry-leading transcription accuracy in 99 languages. And this is now available in the ElevenLabs API, so we can build fun applications, like for example, this transcription Telegram bot. So, I just recorded this voice message in the Telegram bot, and you can see, you know, not just a second later, we get our transcript back. Now, you might be saying, you know, "You have a really good microphone here, and your pronunciation is, you know, somewhat okay." But, uh, I was in the MRT- For passengers heading to Marina Bay Sands- ... you know, in Singapore. ... please transfer to the CCI line, and alight at Bishan Station. And, you know, this might be a bit more challenging, but, so what we can do is we can forward this to our bot. And so now if we send this off, you can see here, "For passengers heading to Marina Bay Sands..." You know, a lot of people might have not understood that, but the model is really, really good in, um, you know, understanding that and accurately transcribing it. Um, another thing we can do is we have the launch video here, for example. So we can also forward videos to this bot because we can actually, uh, put entire videos to the API. And you can see again fairly quickly, so this was kind of a minute-long, um, video here, a bit of a larger file. And you can see also we get some, um, audio cues here, so upbeat music, um, things like that. Now again, you know, this was just English, but we support, um, you know, 99 languages with high, high accuracy. So for example, I'm German myself, so maybe let's give that a go. (German) So I was speaking specifically very, very quickly here, and you can see. So we can speak German here, and I was just kind of testing, okay, uh, if we're speaking quickly. So that's really nice. Uh, maybe we try French. (French) So I was just, uh, saying that, um, my French is very bad, I'm German. (laughs) There we go. It understood that. Or we can try my, uh, broken Mandarin. (Mandarin) And luckily the model here is, um, quite friendly because it actually understands my broken Mandarin. (laughs) So yeah, you can see here, um, really, really cool, really, a lot of use cases we can do with this. So let's have a look at how we can build, uh, exactly this transcription bot with the ElevenLabs Scribe API.