With that clear and concise explanation paired with a single jupyter notebook to make things even simpler, you got my like and subscribe. Thank you and more power!
Oh wow, thank you so so much for this video! I am currently doing my master thesis and I have to transcribe a lot of interviews. This is cutting down the work by a lot! I had some troubles initially with installing brew and ffmpeg, since I didn't know how to do it, but you mentioned it in other videos and it worked. For a beginner like me, coding feels (quite) a bit like magic so it is always exciting when things do work out and I even more if understand what is happening 😄all thanks to you!
@@NicholasRenotte I am taking a closer look on collaborative innovation processes 😊I hope to find out how innovations emerge in cluster focus groups and how we can optimize the innovation process in a cluster setting. Let's see what comes out of this 😁🤞🏼
Hello. What suggestion can you give me for this error: TypeError: request() got an unexpected keyword argument 'continuous'. I have followed all the steps as described in the video. If I leave out the continuous=type expression it stops showing the error. Thank you.
Love your videos man, you deserve to blow up soon! I´ll literally watch all the Python ones. I'll use the Watson for my school video. Btw maybe you could make a video for Python / AI beginners?
Thanks 🙏 so much @Kerim Al-Lami! Python Basics for data science is on the list! I’m going to put it together in a bit of a playlist, what do you think?
Yo thanks so much!! Let's break it down into three steps: 0. Inside of our folder we start with a file called audio.wav, this is a long audio file that is uncompressed 1. This ffmpeg command takes that file called audio.wav (ffmpeg -i audio.wav) and converts it into another file called audio.mp3 (-vn -ar 44100 -ac 2 -b:a 192k audio.mp3) command = 'ffmpeg -i audio.wav -vn -ar 44100 -ac 2 -b:a 192k audio.mp3' 2. The second command then picks up the compressed file called audio.mp3 (ffmpeg -i audio.mp3) and splits it out into 360 second chunks (-segment_time 360) command = 'ffmpeg -i audio.mp3 -f segment -segment_time 360 -c copy %03d.mp3' 3. Then inside of our directory we end up with a number of split files in the format 00#.mp3 (-c copy %03d.mp3). Here's a little more on how ffmpeg works: ffmpeg.org/ffmpeg.html#Description Let me know if you want me to delve a little deeper or explain anything else, I'm always happy to help!!
Actually, I did the research and I found the problem: When you use Win 10, you should add the path of your FFmpeg installation. A bit tricky, but there are tons of videos about the subject. At last, the program works, but the Spanish models didn't recognize much of the meeting I've recorded.
Hello, thats a nice video but i have a problem, in the compress and split audio part, the first code give me a out: 127, I dont know how to do, the name of the file is correct and i installed correctly the ffmpeg, what can i do?
Love all your work...you are great. Question is it possible to create an app that searches audio files for keywords....Like if I want to search through a bunch of my audio files for a certain subject that was discussed is that possible.
Hi, really helpful video ! I am working on a long audio file in Greek and i want to get a dictionary of 20 words from it. Is there any variation i can use in order to extract only the intervals of these words ?
Hi! Great video! I'm not practical with Python but with your lesson I understood how to use this functionality. :D I have a question for you: I'm trying to make text from an Italian speaker lesson of mechanical engineerig and following your notes I arrived at the end but The accuracy of the text is quite poor. To understand, if I use the voice digitization of google doc (by sending the audio in the microphone with a virtual cable) the accuracy is better than Watson STT, but Watson STT is much more comfortable sinse the process is automated.How I could solve the problem of the accuracy? How is possible that the voice digitization in (free) google doc give me butter result than Watson STT? Thanks in advice @Nicholas ! :D
@@NicholasRenotte Yes Yes I'm already using the Italian model. A the end the software works but it is not soo accurate... meaby the italian model is in general not performing like the english one? Thank you soo much bro! :D
@@antoniogrieco9954 ah got it! You can also fine tune the model to improve the accuracy: developer.ibm.com/technologies/artificial-intelligence/patterns/customize-and-continuously-train-your-own-watson-speech-service/
Thank you Nick for a video. but i'm stuck in files part. I suppose i should have indicate a directory at the beginning? update: I have figured out with the directory now i'm stuck with converting wav to mp3.
@@NicholasRenotte Awesome man! THanks for answer. I solved yesterday a problem by watching your other video you are amazing bro! Now I wonder how could I get the sound from the mic directly.
Hi this was very useful !! I have an audio file where two speakers are conversing and I need the sentences to be segregated with the identification as speaker1 and speaker2. Can you please help me for this logic? Thanks in advance.
@@NicholasRenotte Nick I already figured it out!! I was missing the ffmpeg so the code didn´t run completly. But thanks!! It really worked!! Hugs from Colombia!!
@@NicholasRenotte I think I found the issue. When I execute that cell Terminal outputs this: /bin/sh: ffmpeg: command not found I just don't know how to solve it. I installed ffmpeg through pip, because Homebrew isn't compatible with my Mac. Do you know how I could fix it?
@@NicholasRenotte I think it is because ffmpeg is being executed from the shell. Is there way to execute from where it actually is, or move ffmpeg to bin/sh where it should be?
@@NicholasRenotte Real time as in - let's say we want to generate a good summary of any famous personality using the information we have on internet directly without explicitly storing it in the database...(like automatic generation of Wikipedia paper)...It will be really helpful 😀
@@pranavdange5695 ohhhhh got it! Interesting, you could do it with something like a web-scraper and then pass it through a summarization pipeline like this: th-cam.com/video/TsfLm5iiYb4/w-d-xo.html
Hi Nicholas, Thank you so much for sharing this video. I have been trying to transcribe audio files(the duration of each audio file in the folder is 25 seconds). However, the IBM speech to text API is not transcribing all the files. I always get this error: ApiException: Error: Forbidden, Code: 403. What do you think could be the cause of the error and how do I fix it? The total number of speaker in the folder is about 2172
Heya @Luqmon, has the code worked before? A 403 error indicates that perhaps you weren't able to authenticate successfully. Was there a larger error you can share?
@@NicholasRenotte , Yes, I inspect the the error and I think the error is coming from the recognize function of the speech to text service. Also, I saw this too: raise ApiException(response.status_code, http_response = response exxept requests.exceptions.SSLError)
@@NicholasRenotte Hey hello how can i try with a audiofile in spanish model='es-ES_NarrowbandModel' this is correct or i need another thing can you help me?
sorry, I've problem with this part: import subprocess import os command = 'ffmpeg -i audio.wav -vn -ar 44100 -ac 2 -b:a 192k audio.mp3' subprocess.call(command, shell=True) command = 'ffmpeg -i audio.mp3 -f segment -segment_time 360 -c copy %03d.mp3' subprocess.call(command, shell=True) The output is 1. I've windows 10 and I use jupyter notebook like you, what is the problem?
@@NicholasRenotte The problem is that I didn't know I had to install ffmpeg, being at the basics with Python I assumed that if you hadn't called it at first with the import command it would have been a native Python function but I was wrong. I watched your other videos and then went to the official site to install the package. Once installed everything was fine. I am only sorry that all used scripts are not useful to me, unfortunately the recognition in Italian is not as excellent as that in English and I cannot transcribe my university lectures. But I learned something new thanks to you!
greeting bro I hope you are good I wanna read convert file speech to text its about 15 minutes long and I got error please help me this is the error Code : 503 I hope you understand the complete error because I didn't write here complete its the last sentence of error
@@NicholasRenotte Your works are really appreciable.. How to transcribe large Persian and Urdu recorded audio file..?? I really need a transcriber for my lectures please help..?? God will bless You..
Thanks so much @Diego Caumont! Weird, are you getting an error? You should be able to leverage the 500 minute limit. If it's still blocking try deleting the service and recreating it and shoot me the error/code here and I can help debug!
this has saved me so much time at work, Everytime i have to type conferences calls i come to this video. It works very well and its free
With that clear and concise explanation paired with a single jupyter notebook to make things even simpler, you got my like and subscribe. Thank you and more power!
Oh wow, thank you so so much for this video!
I am currently doing my master thesis and I have to transcribe a lot of interviews. This is cutting down the work by a lot! I had some troubles initially with installing brew and ffmpeg, since I didn't know how to do it, but you mentioned it in other videos and it worked. For a beginner like me, coding feels (quite) a bit like magic so it is always exciting when things do work out and I even more if understand what is happening 😄all thanks to you!
Woah, awesome stuff @Kalina, great to hear it's saving you a bunch of time! What's your Masters thesis on?
@@NicholasRenotte I am taking a closer look on collaborative innovation processes 😊I hope to find out how innovations emerge in cluster focus groups and how we can optimize the innovation process in a cluster setting. Let's see what comes out of this 😁🤞🏼
@@kalina8877 fascinating, wishing you all the best with it, hopefully some awesome insights!!
thank you so much - you just saved me a bunch of time transcribing interviews
YESS! Awesome!
I've been searching for a channel like this!
What's happening, welcome to it 😀
that's cool bro! thank you very much for the video, it really helped me out of the confusion of the program I made
YESSS! Awesome!
Hello.
What suggestion can you give me for this error:
TypeError: request() got an unexpected keyword argument 'continuous'.
I have followed all the steps as described in the video. If I leave out the continuous=type expression it stops showing the error.
Thank you.
Love your videos man, you deserve to blow up soon! I´ll literally watch all the Python ones. I'll use the Watson for my school video. Btw maybe you could make a video for Python / AI beginners?
Thanks 🙏 so much @Kerim Al-Lami! Python Basics for data science is on the list! I’m going to put it together in a bit of a playlist, what do you think?
Amazing video! That helped me all the way through I was struggling before using only the API docs! 👍🏽
So glad it helped @ChrisGuitarTV!
thank you so much for this video. I was see your videos. This will be help full to me for put first step to the Automation.
Can you make a video about clearing audio from noise for speech to text please? Especially using algorithm fastICA
Btw i love all your video
good experience for newbie as me. Thank you !
Awesome tutorial, but I'm a bit lost with the compress and split part. There is a preprocess for the audio before? I didn't get how ffmpeg works.
Yo thanks so much!! Let's break it down into three steps:
0. Inside of our folder we start with a file called audio.wav, this is a long audio file that is uncompressed
1. This ffmpeg command takes that file called audio.wav (ffmpeg -i audio.wav) and converts it into another file called audio.mp3 (-vn -ar 44100 -ac 2 -b:a 192k audio.mp3)
command = 'ffmpeg -i audio.wav -vn -ar 44100 -ac 2 -b:a 192k audio.mp3'
2. The second command then picks up the compressed file called audio.mp3 (ffmpeg -i audio.mp3) and splits it out into 360 second chunks (-segment_time 360)
command = 'ffmpeg -i audio.mp3 -f segment -segment_time 360 -c copy %03d.mp3'
3. Then inside of our directory we end up with a number of split files in the format 00#.mp3 (-c copy %03d.mp3).
Here's a little more on how ffmpeg works: ffmpeg.org/ffmpeg.html#Description
Let me know if you want me to delve a little deeper or explain anything else, I'm always happy to help!!
Actually, I did the research and I found the problem: When you use Win 10, you should add the path of your FFmpeg installation. A bit tricky, but there are tons of videos about the subject. At last, the program works, but the Spanish models didn't recognize much of the meeting I've recorded.
@@NicholasRenotte Thanks for all the help
Hello, thats a nice video but i have a problem, in the compress and split audio part, the first code give me a out: 127, I dont know how to do, the name of the file is correct and i installed correctly the ffmpeg, what can i do?
same on my end
i ended up creating a virtualenv and brew installing ffmpeg in that environment, as pip install ffmpeg wasn't working. all good now!
Heya @Albertl, just checking, have you double checked your file paths?
Love all your work...you are great. Question is it possible to create an app that searches audio files for keywords....Like if I want to search through a bunch of my audio files for a certain subject that was discussed is that possible.
Yup! I think one of my colleagues at work did something similar for TEDx videos!
@@NicholasRenotte could you share a link to this please?
Hi, really helpful video ! I am working on a long audio file in Greek and i want to get a dictionary of 20 words from it. Is there any variation i can use in order to extract only the intervals of these words ?
Is there a way to add the correct capitalisation and commas, full stops etc?
Could do it manually, or use a grammar based DL model?
Hi! Great video! I'm not practical with Python but with your lesson I understood how to use this functionality. :D I have a question for you: I'm trying to make text from an Italian speaker lesson of mechanical engineerig and following your notes I arrived at the end but The accuracy of the text is quite poor. To understand, if I use the voice digitization of google doc (by sending the audio in the microphone with a virtual cable) the accuracy is better than Watson STT, but Watson STT is much more comfortable sinse the process is automated.How I could solve the problem of the accuracy? How is possible that the voice digitization in (free) google doc give me butter result than Watson STT? Thanks in advice @Nicholas ! :D
Heya @Antonio, just wanted to check you're already using the Italian model?
@@NicholasRenotte Yes Yes I'm already using the Italian model. A the end the software works but it is not soo accurate... meaby the italian model is in general not performing like the english one? Thank you soo much bro! :D
@@antoniogrieco9954 ah got it! You can also fine tune the model to improve the accuracy: developer.ibm.com/technologies/artificial-intelligence/patterns/customize-and-continuously-train-your-own-watson-speech-service/
is there a way to make this into an app or something that I could, for example have my friend download to use on his pc?
hey why do we break up the audio in parts?
API has a limit for chunked requests.
Thank you Nick for a video. but i'm stuck in files part. I suppose i should have indicate a directory at the beginning?
update: I have figured out with the directory now i'm stuck with converting wav to mp3.
Heya @Ibrahim, here to help, what's the error when converting the wav to mp3?
@@NicholasRenotte Awesome man! THanks for answer. I solved yesterday a problem by watching your other video you are amazing bro! Now I wonder how could I get the sound from the mic directly.
@@ibrahimisrafilov1248 awesome work! You can do it as a stream, I think I've got the code somewhere let me know if you need it!
@@NicholasRenotte have looked on your Git Rep. but didn't find the one I need more. It could be so cool if you could share with me.
@@ibrahimisrafilov1248 check this out: github.com/IBM/watson-streaming-stt
Hi this was very useful !! I have an audio file where two speakers are conversing and I need the sentences to be segregated with the identification as speaker1 and speaker2. Can you please help me for this logic? Thanks in advance.
Heya @Vinay, don't have anything in that space yet unfortunately!
@@NicholasRenotte That is completely ok !! Thanks for replying !! Thanks for the great stuffs though !!
@@svk-creatives anytime! Thanks for checking it out!
Hello Nick! I have a problem loading my audio files to Jupiter. I can´t find them and when I call them they don´t appear in the code. What can I do?
Heya @Maria are your audio files in the same folder as your Jupyter Notebook?
@@NicholasRenotte Nick I already figured it out!! I was missing the ffmpeg so the code didn´t run completly. But thanks!! It really worked!! Hugs from Colombia!!
@@mariaceciliagarciagutierre2390 awesome work!! Right back at you from Sydney!
Hmmm when compressing and splitting audio mine doesn't return a 0 but a 127 and it doesn't split the audio. Any idea about why that may be?
Hmmm, weird, what file types are you using?
@@NicholasRenotte The file is originally a .m4a. I tried turning it into a .wav and it still didn't work. Maybe I changed it wrong?
@@NicholasRenotte Converted the file to wav through Audacity, but still no luck. It still outputs to 127
@@NicholasRenotte I think I found the issue. When I execute that cell Terminal outputs this:
/bin/sh: ffmpeg: command not found
I just don't know how to solve it. I installed ffmpeg through pip, because Homebrew isn't compatible with my Mac. Do you know how I could fix it?
@@NicholasRenotte I think it is because ffmpeg is being executed from the shell. Is there way to execute from where it actually is, or move ffmpeg to bin/sh where it should be?
How to transcribe from the stream of the audio in real time whisper AI model?
Bro, amazing content...
Can you please make a real time text summarizer using Deep Learning
Real time, as in from speech?
@@NicholasRenotte Real time as in - let's say we want to generate a good summary of any famous personality using the information we have on internet directly without explicitly storing it in the database...(like automatic generation of Wikipedia paper)...It will be really helpful 😀
@@pranavdange5695 ohhhhh got it! Interesting, you could do it with something like a web-scraper and then pass it through a summarization pipeline like this: th-cam.com/video/TsfLm5iiYb4/w-d-xo.html
Hey Nic, thanks for posting! Question for you, or anyone who sees this comment. Any luck on using ffmpeg to convert .M4A to .MP3 or even to .wav?
hahaha Nick that beginning.
Can we have something on Computer Vision, please?
🤣 I had too much fun making it! Definitely, real time mask detection is coming this week! Any other CV stuff you're thinking of?
Hey @Aman Rathi, I think we had a chat about some more CV stuff, check this out: th-cam.com/video/IOI0o3Cxv9Q/w-d-xo.html
@@NicholasRenotte WOW... I am so happy we finally have that... So excited to start this, 58 minutes is really good... THANKS NICK
@@mummyskitchen5311 anytime man!! Let me know how you go!
where is the audio file saved? no need to give file path?
Should be saved into the current directory @Aravind!
Hi Nicholas, Thank you so much for sharing this video. I have been trying to transcribe audio files(the duration of each audio file in the folder is 25 seconds). However, the IBM speech to text API is not transcribing all the files. I always get this error: ApiException: Error: Forbidden, Code: 403. What do you think could be the cause of the error and how do I fix it?
The total number of speaker in the folder is about 2172
Heya @Luqmon, has the code worked before? A 403 error indicates that perhaps you weren't able to authenticate successfully. Was there a larger error you can share?
@@NicholasRenotte , Yes, I inspect the the error and I think the error is coming from the recognize function of the speech to text service. Also, I saw this too: raise ApiException(response.status_code, http_response = response exxept requests.exceptions.SSLError)
Heya @@lukmonafolabirasaq860, quick check, what URL are you using?
@@NicholasRenotte I am using this url: api.us-east.speech-to-text.watson.cloud.ibm.com/instances/c1c13552-e4ce-4b85-bff6-fd45ba2c1006
@@NicholasRenotte Hey hello how can i try with a audiofile in spanish model='es-ES_NarrowbandModel' this is correct or i need another thing can you help me?
sorry, I've problem with this part:
import subprocess
import os
command = 'ffmpeg -i audio.wav -vn -ar 44100 -ac 2 -b:a 192k audio.mp3'
subprocess.call(command, shell=True)
command = 'ffmpeg -i audio.mp3 -f segment -segment_time 360 -c copy %03d.mp3'
subprocess.call(command, shell=True)
The output is 1.
I've windows 10 and I use jupyter notebook like you, what is the problem?
nothing, I solved
Awesome work @Matteo, what was it?
@@NicholasRenotte The problem is that I didn't know I had to install ffmpeg, being at the basics with Python I assumed that if you hadn't called it at first with the import command it would have been a native Python function but I was wrong. I watched your other videos and then went to the official site to install the package. Once installed everything was fine. I am only sorry that all used scripts are not useful to me, unfortunately the recognition in Italian is not as excellent as that in English and I cannot transcribe my university lectures. But I learned something new thanks to you!
@@MatteoGriot ah, got it, noted for next time! Awesome stuff!
Can you please make a video on how to create Speech To Text Machine learning Model building?
Like from scratch?
@@NicholasRenotte Yes Sir... From Scratch.... Without using any 3rd party API or anything......
@@a2sirmotivationdoses782 will add it to the list, might be a little while coming though!
@@NicholasRenotte OK Sir thanks for the Support and Can I get your WhatsApp number please?
@@a2sirmotivationdoses782 anytime!
greeting bro I hope you are good I wanna read convert file speech to text its about 15 minutes long and I got error please help me
this is the error
Code : 503
I hope you understand the complete error because I didn't write here complete its the last sentence of error
Shoot me the error, let's take a look!
It does its job, but for some reason it only creates the 000.mp3 and that is the only one it coverts. (coding in pycharm)
Heya @jonas, is your file smaller than the chunk size?
Hey, that lecture hall looks familiar, what uni do you go to?
Photoshop Technical University 🤣 !
@@NicholasRenotte hahahah
where do you actually put the file in?
Heya @Saurav, in the same folder as your Notebook.
hey what if im planning on transcribing non-english audios, will it work also???! thankyou, your videos have been very helpful
my audios are in Indonesian
Heya @GlennD, unfortunately it doesn't look like Indonesian is supported: cloud.ibm.com/docs/speech-to-text?topic=speech-to-text-models
@@NicholasRenotte can you do one using google speech to text API???
@@NicholasRenotte Your works are really appreciable..
How to transcribe large Persian and Urdu recorded audio file..??
I really need a transcriber for my lectures please help..??
God will bless You..
@@mohammedmehdi1248 Hmmm, I don't think those languages are available on the WSTT API @Mohammed, my bad man.
how about speech diriziation ?
Sort of like this? th-cam.com/video/8k8S5ruFAUs/w-d-xo.html
Dope, as always :)
Just a quick question, how come am I only limited to a 5120byte file? In comparison to your 500 minutes limit...
Thanks so much @Diego Caumont!
Weird, are you getting an error? You should be able to leverage the 500 minute limit. If it's still blocking try deleting the service and recreating it and shoot me the error/code here and I can help debug!
can i use ur apikey and url ?
Is this Available in every language..???
Nope, but a bunch are supported: cloud.ibm.com/docs/speech-to-text?topic=speech-to-text-about
@@NicholasRenotte bro Can you make a code for me for transcribing in Persian language because I really need it for my University lectures..
what is ffmpeg ? you dont tell us to installl...?????
ahh!!.. i get that..!!! thank you
Perfect! So glad you got it working man!
sir how can convert speech to gesture ??
Take a look at GANs!
Please write the all codes..
just built a time saving machine,,
Getting this error message. Any idea what this means? TypeError: Session.request() got an unexpected keyword argument 'continuous'
Yes same for me too . Hope someone helps us 😭