Speech and Lyrics Transcription with AI

Hello and welcome back.

Today we're going to talk about DaDaScribe.

Dadascribe.com is a website we have developed to help with speech transcriptions.

If you look at this video or all the videos we have on our website, we have a textual transcription of the video itself.

Those transcriptions have been made initially by real human beings, but lately from this new system which is of course AI based, which make the transcription automatically for us.

The system is amazing because it's very cheap, it's very fast, it's very precise. I'd say that it's comparable to a transcription made by a real human being, so you should really try it. The funny and interesting thing is that we have noticed that people are using the same system even just to transcribe lyrics from songs, which is crazy. We really didn't think about that. People could use it for doing that. It actually has not been designed to do that. Sometimes you screw up things or transcribe wrongly the lyrics from the songs, but you could try to give it a try because it's not that bad. I would say probably 90 -95 % of the times it's correct.

Give it a try and let us know. In the meantime, I'm going to show you in this video how Dadascribe works and how it can help you with speech transcription. If you want to try with the lyrics, you can try that as well. Alright, so let's dive right in.

If you head to dadascribe.com you will find the homepage which will allow you to start transcribing in three different ways. You can either enter our YouTube URL or a link or upload an audio file or record. You can also record your own voice live.

In this demo I'm going to select YouTube and I'm going to try to transcribe one of our videos, like the one I did the last time. I'm going to take the link from here, copy the link, get it back to Datasccribe, paste it inside and I'm going to click Next.

Now the system is going to check the URL to make sure it is correct and valid.

Now it's asking me if it includes multiple speakers because you can also transcribe speeches between more than one person. In this case it's just me so I'm going to click No and it's going to ask for the language. This system can transcribe in hundreds of languages, in this case English. I'm going to click Next.

It can also make translations in up to five different languages. I'm going to just keep English in this case and maybe I'm going to try it to Italian and see how it works.

I'm going to click Next and now it's done. I go into my transcription panel. As you can see here my request has been queued and you will see that really soon it's going to start transcribing.

This video is seven minutes long and the transcription will take probably less than a minute.

I'm going to wait until it starts and I'll be back as soon as this is done.

Alright, it's complete. It took a little bit more than one minute. I'm going to click on the transcription and as you can see we have the source video at the top and we have our transcription here which is probably 99 % accurate if not 100%.

We also have subtitles so we can upload them into YouTube if you want to another video editing program to have subtitles ready to be used.

Also here is my Italian translation which is pretty good.

It looks really really well done and subtitles of the Italian translation. So this is basically the service but let's try with a song as I said. Let's try with... Let's start over. Let's try with the famous Careless Whisper by George Michael. I'm going to take the link of this video.

I'm going to get back to Dadascribe homepage, enter it into the YouTube URL or link field and let's start the process again. I'll see how it works.

Okay so no multiple speakers, source language English and yeah let's try to... this time let's do French and see how it works.

Okay let's go. Now we're gonna go here and wait until it's done. This is five minutes long so I expect to take less than a minute.

All right this is done so let's see how it went.

Yeah it looks pretty good.

Maybe some word is not perfect or I missed something but it works to some extent and this is the French transcription.

Okay so as I said it's not perfect for songs because it wasn't designed to transcribe songs but it can be used to some extent and of course it depends also by material you have. If you try to transcribe heavy metal it's probably going to give you almost nothing because it's too noisy but for very clean songs it could work 90 % up to yeah I think 90 % accuracy.

And how does it work? I mean if you go on the about page of Datasccribe the system just cleans the source from background noise or music leaving just the voice track and that's the way it works.

And then it uses an AI based system called Whisper which is by OpenAI which is open source actually and the way it works really well. And also there is the feature that if you have multiple speakers you can divide the voices with a kind of process which is called the diarization and of course the last process the optional process is translation. You can actually translate the source to over 120 languages which is amazing.

Anyway let me know what you think and if you try the system you can of course try it for free for I think 10 minutes and let me know how it works for you.

Okay, so what do you think about this Dadascribe? It's pretty interesting.

It's not really music focused but it's interesting because it can be used to transcribe videos about music as well or lyrics if you want to try it and maybe you can adjust it manually the transcription but you can try with that. Anyway let me know if you have any questions post your comments below and get in touch with me as usual through the email tips and insights at virtualsheetmusic.com and I'll see you in the next one. Bye bye.

Automatic video-to-text transcription by DaDaScribe.com

Thank you for your comment, question or request!

Comment/Question deleted successfully.

Post a comment, question or special request:

Comments or questions must be between 3 and 10,000 characters long.

You may: Login  or  

Otherwise, fill out the form below to post your comment:

Add your name below:


Add your email below: (to receive replies, will not be displayed or shared)


For verification purposes, please enter the word MUSIC in the field below

---