Speaker 1: Hi everyone, I'm Patrick, I'm a developer advocate at Assembly AI, and in this video I want to show you how you can very quickly try out all our powerful AI models in just a few seconds and with them transcribe and analyze audio content. So for this you don't need an account and you also don't need to code. You can use our Assembly AI Playground that you find at assemblyai.com slash playground. And here you can upload any local audio or video file, or even better, you can drop in any YouTube video link. So let's get started. So let's quickly test our AI models with a YouTube link. And for this, you could even click on surprise me and it's randomly choosing one, I want to use one from the Kurzgesagt channel, which is this video, what happens if a super volcano blows up. So let's copy the link. And then let's paste this in here. And now click on Next. And here you get an overview of all the models that you can try out. By default, transcription is always turned on. And now in addition, I also want to select summarization. So this can generate a single abstractive summary of the entire source submitted for transcription. And now we come back to this overview again in a few moments and have a look at the other ones as well. For now, I want to go to next and now it's already starting this and running this through our models. Processing time takes roughly 15 to 30% of the file's duration. So we need to come back in a few moments. And in the meantime, we can copy the URL and leave the site. So let's wait until the results are here. And we got the results back on the left side, we now see the transcript of the whole video. For example, let's read the first sentence, the earth is a gigantic ball of semi molten rock with a heart of iron as hot as the surface of the sun. This is exactly what's being said in the video. And on the right side, we see the summarization. In this example, we got two bullet points back. So for example, let's read through some of this true apocalypses can break through and unleash eruptions 10s of times more powerful than all of our nuclear weapons combined. So yeah, I'm a big fan of the summarization model. This is super cool to get a very quick summary of the whole video. And now let's have a look at the other AI models you can test out as well. So let's use the same YouTube video again and go to the model overview. We've just seen the summarization. Now let's also use topic detection. So this can label the topics that are spoken in your audio or video files following the standardized IAB taxonomy. Then we can turn on all the chapters, this will generate a summary over time, providing an automatically generated summary for each chapter, then we can use content moderation. So this will detect any sensitive content, we can detect important phrases, we can do sentiment analysis, which will detect the sentiment for each sentence. So either positive or negative. We can detect entities such as person or company names or locations, then we can do PII reduction. So personally identifiable information, such as phone numbers and social security numbers, we can detect and label speakers, we can also have dual channel support. So if we have a dual channel audio, then this will transcribe each channel separately. And we can also turn on profanity filtering. So in this example, let's turn on these here topic detection, auto chapters, content moderation, important phrases, sentiment analysis. And let's go to next. And now let's wait again until the results are here. And processing is done. Now on the right side, we again get the results for all the AI models that we turned on. So let's go over them. First, we have order chapters. So in this case, it detected two chapters. And for each one, we get a headline and then the summary. So the headlines are could a super volcano destroy the earth and the second one are super volcanoes a threat to humanity. And then the summary, which I think is pretty cool. Then we have content moderation. So I think for a YouTube video, there's not a lot of sensitive content. But in this case, it detected disasters as possibly sensitive content. Then the important phrases, for example, volcanic eruptions, super eruptions, hot rock, climate change and many more. Then for the sentiment analysis, it detected 13 positive and 27 negative ones. So I think this video is talking a lot about these disasters and death. That's why there's also a lot of negative sentiment in here. And for the topics, it detected geology, disasters, polar travel, science, weather and online education. So I think this is pretty accurate. So yeah, I think these AI models are pretty cool. And I encourage you to try them out for free at assembly.com slash playground. And one quick note, if you then work with the actual API and code, of course, the results you get back are much more detailed than what you're seeing here. For example, for the other chapters, you also get the timestamps back for each chapter. And you also have more customization. So for example, you can customize the summarization model. But for this, you can check out the documentation. But I think the playground is super cool to test out any YouTube video or local file really quickly. And yeah, I hope you enjoyed this video. And then I hope to see you in the next one. Bye.