If you are interested in these topics, you can also follow me on Twitter :) twitter.com/terzi_federico

@jeongwonkim247

3 жыл бұрын

was there a video on how to transcribe the audio files into text? Please let me know and thank you!

I really like, that you mention the platform at the end!

Works on Windows after struggling for a while, many thanks

Thank you. Started DeepSpeech in a minutes.

thank you for sharing. i donated my voice there. this is so cool!

@FedericoTerzi

3 жыл бұрын

That's great! :) We are one little step closer to an open voice model

Thanks so much for making this video, it was exactly what I was looking for!

Thanks for this video! I got everything working. As you said, it's not as good as proprietary solutions but for simple commands like "start", "stop" or "turn on light" it should work. Looking forward to contribute.

@FedericoTerzi

3 жыл бұрын

Glad you liked it :)

Thanks for this short, straightforward, to-the-point video! By reading the manual I thought I am going to need to take a vacation to learn just to run DeepSpeach, now I am very confident about doing it quite quickly!

@FedericoTerzi

3 жыл бұрын

Thanks! Running it is pretty easy with the prebuilt model. Things start to get real complex when you want to train your own :)

@KuboF

3 жыл бұрын

@@FedericoTerzi Yeah, using pre-built model is my first step to training my own 😅

@FedericoTerzi

3 жыл бұрын

Good luck! If you succeed, please let me know how hard it was :)

@KuboF

3 жыл бұрын

@@FedericoTerzi I very much hope I could one day 😅

@patataboom2645

3 жыл бұрын

Soo have you finished? :))))

I was looking for this kind of video.. Currently I am working on creating AI assistant. This will help me a lot

realy helpful for me (I think your video is the only one on the subject?). in addition to this, a tutorial on mozilla's TTS would actually be great. I would like something more detailed for that. I currently don't understand how to use new datasets to get other voices. i guess you have to train a model with a dataset. a tutorial on this would be really really cool! maybe you have also dealt with it?In any case, deepspeech and tts can theoretically be combined well.

Great Video, I had some problems but now it works. Thank you

@FedericoTerzi

3 жыл бұрын

Thanks!

Thanks for ur video! I used Python 3.8.6 and DeepSpeech v0.9.3 and it worked well!

@hssp1534

Жыл бұрын

but im not able to find the deepspeech library in jupyter. How did you install it?

@jane_shi

Жыл бұрын

I just did what he showed in the video

Great video. Thank you 😊

Job nicely done and presented, thank you.

@FedericoTerzi

3 жыл бұрын

Thanks!

Thanks for making this man!

@FedericoTerzi

4 жыл бұрын

You're welcome :)

Uuu, really good content here! Grandee

@FedericoTerzi

4 жыл бұрын

Thanks :)

Got this working on windows! thanks for the tut!

@FedericoTerzi

3 жыл бұрын

Glad to hear that :)

@sauravprashar

3 жыл бұрын

Could you please help me I am getting a DLL error

@LukeHildreth

3 жыл бұрын

@@sauravprashar I'm actually not sure how to answer that. I'm pretty new to programming. Hope you find the answer!

Wow great library thank you

Excellent. If you could make a tutorial on how to train own model. The big players have already done that for well-known languages. In contrary this one could help with neglected languages like mine. So a tutorial on how to train own model in a new language would be very helpful.

@FedericoTerzi

3 жыл бұрын

Thank you! Unfortunately, I don't know the model that well...

super nice! was able to follow it successfully, but on a mac.

@FedericoTerzi

4 жыл бұрын

Glad to hear that :)

@LukeHildreth

3 жыл бұрын

I'm trying this too. Hid did you activate the script after setting up the virtual environment?

very nice . plz make a video of how you train your own model

Do you think we can run whisper or fast whisper llm on esp32's? Sort of in a form factor like the carputer or beepberry?

Thank you!

Thanks, finally a good tutorial for Deepspeech!

Hi Terzi, can this software run on mac? and can it translate movie videos in to any language? thanks in advance!

can you please do a video about implanting this code to a project please?

Hi Federico, thank you for your video. do you know how to integrate Deepspeech into talon?

does the vad streaming work outside deepspeech? i wanna use it in another ASR framework

Can we somehow play with the energy(loudness) levels under which audio is captured , triggering the transcription.? I mean every little sound triggers deepspeech, while we want it to be triggered only when a person speaks. Can somehow an energy threshold be determined? I'm working on it, but I could save some time if there is already a solution. There seems to be a condition in vad_collector() that I am finding hard to understand. Thank you

@FedericoTerzi

3 жыл бұрын

Hey, yes that's almost surely possible by playing around with the audio stream. I don't know exactly how though

Hi can I ask on how I can integrate this to my application? My application is now working on ionic environment is it possible to integrate this one?

Do you can do a tutorial teaching about how use deepspeech in unreal engine?

What if I have multiple wav files and I want to transcribe each one and output to a text file? How do I do that?

Thank you for the video. Nice tutorial to get familiar with!!!!! Also, I have found a german pre-trained model. could you please explain how to work with german or any other pre-trained model.

@FedericoTerzi

3 жыл бұрын

You should be able to simply pass the german model and scorer and you should be ready to go :)

@chaitanyamalpure6226

3 жыл бұрын

@@FedericoTerzi Thanks alot. It worked!!!!!!!!!!!!!!!!!!!!!

Thank you thank you thank you

Nice video. Can you make a video about the use in android?

Hi! You are using an already pre-trained model to do this speech-to-text application. But what if you want to train this model with another dataset, like for example in spanish or in italian? Which would be the steps to take in order to train the model to recognize speech in another language that isn't english?

@ThesongsIlikeThemost

3 жыл бұрын

hi, you can find already trained model for Spanish, Italian, German, Polish, and French here. gitlab.com/Jaco-Assistant/deepspeech-polyglot

@sebastianochipocomancini1853

3 жыл бұрын

@@ThesongsIlikeThemost Thank you so much, I finally found the spanish model here: drive.google.com/drive/folders/1-3UgQBtzEf8QcH2qc8TJHkUqCBp5BBmO (which is a link that was on the url you sent me). Replacing the .pbmm and the .scorer files in the command line, it works fine for spanish!

nice info

was wondering if you could help. I'm getting this error: ERROR: Could not find a version that satisfies the requirement deepspeech (from versions: none) Also, where do you put the two model files after you download

@KPawan108

8 ай бұрын

I am also getting the same error. Did you get the answer now?

Heyy, Pls Solve my query , How many languages does it support ? Like english , spannish could you provide a list of it., I tried to find it on Github and reddit, but was unsucesfull !!!

Very nice video,

@FedericoTerzi

3 жыл бұрын

Thank you :)

How can we build a speech to text model from scratch in other less resourced languages using deepspeech?

thanks !

@FedericoTerzi

3 жыл бұрын

You're welcome :)

Hey, can it be possible upon using colab environment

At 5:04 You told you're going to show nice little project what is it ??

What should I do if I want to use an application like this one for another language like spanish?

@stefang5639

3 жыл бұрын

You can download the language model for other languages as well from the source shown in the video.

ERROR: Cannot install deepspeech==0.9.3 and numpy>=1.15.1 because these package versions have conflicting dependencies. deepspeech 0.9.3 depends on numpy=1.12.0 I am unable to get around this to work, any idea?

Hope we could use it in the Windows environment so that I can study foreign languages easily somemday : )

What should i do if my laptop has installed python 3.8 version

Ciao, Federico. Complimenti per il video, molto bello e interessante! Deep Speech funziona anche per l'italiano?

@FedericoTerzi

3 жыл бұрын

Grazie mille! Si c'è un modello italiano, la performance non è il massimo ma funziona: github.com/MozillaItalia/DeepSpeech-Italian-Model

@Piriponzolo

3 жыл бұрын

@@FedericoTerzi Ciao e grazie, Federico. Ho scompattato lo zip, ma poi mi sono arenato. Come si va avanti?

@FedericoTerzi

3 жыл бұрын

Dopo il processo dovrebbe essere simile a quello del video, anche se non ho mai provato a farlo girare direttamente (ho solo fatto delle prove con il Bot telegram che lo usa). Ti conviene guardare gli esempi sulla repo o contattare il maintainer, che sembra molto preparato a riguardo :)

@Mr_Yod

3 жыл бұрын

@@FedericoTerzi Grazie: lo proverò. Gli altri sistemi che ho provato in Python sono atroci o richiedono la connessione all'internet (quello di google). Certo però che essere compatibile solo con Python 3.6 quando siamo alla 3.9 già da un po'... =( EDIT: Dal link che hai messo dice "Requisiti: 'Python 3.7+' "

good video

Can you use deepspeech for a diacritic system

Awesome!! can it recognize from files too? and how, please?

@FedericoTerzi

3 жыл бұрын

If I'm not mistaken, the script used in the video also accept an argument for wav files :)

Thanks for the video! Is any guidance on how to integrate DeepSpeech into an application on Windows? I'm sure that would be very useful for developers! Thanks!

@FedericoTerzi

3 жыл бұрын

Hey, If you app is written in Python, the integration would be pretty easy. Otherwise, your best bet is to look at "tensorflow-lite deepspeech", although I don't have any experience with that

@robc3863

3 жыл бұрын

@@FedericoTerzi Hi, thanks but our app is C++, but so far not found any example of binding DeepSpeech to it. We also don't have many clients with nVidia GPUs...

@FedericoTerzi

3 жыл бұрын

Nvidia GPUs are really not needed (as long as you are not training the model on the client's PC), CPU will handle inferring ok for most use-cases. Regarding the lack of examples, I'm sorry about that, probably the recent Mozilla layoffs did not help the project...

Speech to text transcribe open source library using python pycharm an another id Es using method code is available bro?

Ciao Federico, grazie per il fantastico tutorial! Mi chiedevo se magari potevi gentilmente potevi farne uno con l Italian Model per i principianti come me

@FedericoTerzi

3 жыл бұрын

Ciao Gianluca, grazie per i complimenti! Non posso prometterti niente dato che non è la mia area di competenza, ma me lo segno :)

@Federico Terzi, Im from India. It is not recognizing Indian English accent. Any thoughts on how to get Deepspeech recognize

@FedericoTerzi

3 жыл бұрын

Unfortunately, there's not much we can do about it. That model was trained on American english, so it struggles with other accents

„Scripts\activate“ is not working. I get a error „parameter wrong -850“

hmmm after I install DeepSpeech into my venv, I do not see "mic_vad_streaming"... any idea on how to install it?

@FedericoTerzi

3 жыл бұрын

You have to download it from the deepspeech examples: github.com/mozilla/DeepSpeech-examples

i already run the code, but it cannot recognize my voice, anyone has solution for it

how did you directly get mic_vad_streaming folder inside the deepspeech folder without doing anything

@abdullamasud4278

3 ай бұрын

he cut out that part from the video. After downloading the file, he simply copy pasted it inside the folder

NIce Tutorial..! So I have a custom trained Language model which has (.model) extenstion, how to I convert it into .scorer file?

@FedericoTerzi

3 жыл бұрын

Thanks! They are two different things, you can't convert one into the other :)

so I followed everything you did. originally started with python 3.7 and it indeed eventually ran int an error trying to install the requirements.txt so I downgraded to 3.6.8. deleted the folder and started over. this time I got everything to work and when i run the mic_vid_streaming.py with the downloaded files, it says "listening..." and whenever I speak it says "Recognized: " but says nothing after that. it clearly is hearing me because it only spits out "Recognized: " when I say something, but then it doesn't print what I said. have any idea what it might be? I'm a begginer to python and coding in general but I was trying to troubleshoot by changing line 194 to text = stream_context to see if my words were somehow in that but it just says "Recognized: " not sure what that means

@FedericoTerzi

3 жыл бұрын

Perhaps it does not hear you loud enough, can you try with another microphone? If I recall correctly, there is a "device" option in the script to specify it

I would like to use deepspeech on a website. Is there a good method for running this in a Javascript environment?

@FedericoTerzi

3 жыл бұрын

There might be, but I dont' have any experience with it! If you are OK only supporting Chrome, then your best bet would be the Web Speech API (which is free and works great): developer.mozilla.org/en-US/docs/Web/API/Web_Speech_API/Using_the_Web_Speech_API

amazing, thank you. can you please make a new video how to install it on debian and train it?

@FedericoTerzi

3 жыл бұрын

Thank you! The installation process should be pretty similar on Debian, as long as you have the right python version. Regarding training the model, that's very difficult and expensive to do...

There is an article on internet how to make srt files for movies via deepspeech. I can't get that working in the windows can you make a video how to convert audio files into text or srt via deepspeech pls? it would be very useful and increase your video views

can this use for movies ?

help Import Error: DLL load failed: no se puede encontrar el modulo especificado

Is deepspeech compatible with nvidia Rtx3090 ?

I get stuck at installing the requirements.txt ERROR: Could not find a version that satisfies the requirement deepspeech~=0.8.0 ERROR: No matching distribution found for deepspeech~=0.8.0 Edit: I already downgraded to Python 3.9.0 which is said to be compatible...

what is the difference between deep speech and deep speech 2?

Hello Federico, Thanks for the video. I followed the steps but after recognizing the audio, it is not returning any result in the command line. After "Recognized: " it's just blank. Sorry, I can't attach any screenshot. I'm using deepspeech 0.7.4 and it is working fine with audio to text conversion. But for microphone application throwing this issue.

@FedericoTerzi

3 жыл бұрын

Hi, most probably you have a problem with your microphone, check if you can record correctly using another application

@fablefoxweaver

2 жыл бұрын

I have the same problem. It listens fine (as in, in detects that I'm talking), but doesn't recognize any words. Using 0.9.3.

(deepspeech) C:\Users\user\Documents\deepspeech>pip3 install deepspeech ERROR: Could not find a version that satisfies the requirement deepspeech (from versions: none) ERROR: No matching distribution found for deepspeech pip and pip3 not working on version 3.10 any ideas?

hi , i cant install requirement.txt with pip . can you help me

I'm working on a college project and I need to make the speech-to-text in my language. Any idea how to use deepspeech in Romanian? I saw the language is available

@mozes_ma

2 жыл бұрын

Hey, similar challenge here, any ideas so far?

is this only with english langauge?

@FedericoTerzi

3 жыл бұрын

You can search online for other pre-trained models, try to google "Deepspeech model"

How much work and time does it need to create a library for new language with its own rules which are not in these libraries?

@FedericoTerzi

2 жыл бұрын

A lot of time, effort and computational power :) You might also want to check out Vosk alphacephei.com/vosk/models

@mo9204

2 жыл бұрын

@@FedericoTerzi is there tutorials for creating own model and training?

Awesome video ! i have already following this instruction the step one by one, but at the end, when the last step. i have a problem "ModuleNotFoundError: No module named 'webrtcvad' . can you help me with this problem? thanks.

@FedericoTerzi

3 жыл бұрын

You might need to install the package with: "pip install webrtcvad"

@fahrul8025

3 жыл бұрын

@@FedericoTerzi Awesome! It works now. But when im speaking there is no any results. Im sure my microphone works well. Any suggestion?

hello i install everything like you but finally the message of missing dll appere .. ImportError: DLL load failed: Une routine d’initialisation d’une bibliothèque de liens dynamiques (DLL) a échoué. can you help me please thanks

@at-ro9217

3 жыл бұрын

same issue here

@FedericoTerzi

3 жыл бұрын

Hey guys, try with these steps: github.com/tensorflow/tensorflow/issues/23683#issuecomment-532522740

Is it possible to write these commands into a python file and just run that?

@FedericoTerzi

3 жыл бұрын

Sure! You can simply edit that script file to fit your needs :)

can you please make a video how to train ?

Interesting topic. I expected to use it in my project but when I tested, it doesn't regognize my voice as good as your :( It catches 50% of my words, and most of the time, it writes a wrong one :( But thank you anyway!

@FedericoTerzi

3 жыл бұрын

Yeah, I've experienced the same problem myself. The model is not comparable with cloud-based solutions as of now, especially for non-native speakers like me :)

was there a video on how to transcribe the audio files into text? Please let me know and thank you!

@FedericoTerzi

3 жыл бұрын

Yes, you can use the script to transcribe audio files as well, but be prepared for some not-so-good results. What's the script "--help" option

Can I use it to generate Bengali language?

Hello! How DeepSpeech can be compared with Open AI Whisper?

@FedericoTerzi

3 ай бұрын

No reason not to use Whisper today! It's amazing

Different numpy versions requirements make it fail for me. deepspeech 0.9.3 numpy 1.14.4 pip 10.0.1 PyAudio 0.2.11 scipy 1.5.4

What should I add to change the voice recognition language?

@FedericoTerzi

3 жыл бұрын

If you are lucky, you might be able to find a pretrained model for your language online. At that point, you can simply point the script to the other model. If you can't find it, then you could create your own model in theory, but that is very difficult in practice

Can you also Integrate deepspeech into a web app with some API? thanks

@FedericoTerzi

2 жыл бұрын

If you're using Chrome or Safari, you might want to check out the Web Speech API, which is much simpler for web apps :) developer.mozilla.org/en-US/docs/Web/API/Web_Speech_API

how can i download mic_vad_streaming

I have no skills in Python, I'm trying to read from wav file instead of mic and display metadata, I tried -f arg but didn't work :( any help ?

@tamgaming9861

3 жыл бұрын

I havent got it to work because i cant install python3.6, my python is already higher. But what i read is that you need a special version of wav-format. I mean to remember it was 8 bit, and mono and 16khz but not sure. MP3 does not work so far. There are some softwares who can translate from mp3 to wav online. Hope it helps.

IBM Via Voice worked perfectly already in 1998, I wonder what happened to it? With IBM Via Voice you could speak and the text was fed into text editor.

@FedericoTerzi

3 жыл бұрын

There are a ton of great (commercial) speech to text products out there. The biggest selling point of DeepSpeech, even though it doesn't perform as well as commercial alternatives, is that it's opensource and free to use, which opens up a ton of possibilities by itself! Unfortunately, the future for DeepSpeech is uncertain at the moment, as Mozilla is cutting all non-essential projects...

@tamgaming9861

3 жыл бұрын

@@FedericoTerzi can you make a tutorial for python3.8 or higher? I cant downgrate python and the higher deepspeech versions have different filetypes now. would be awesome if you can show also how to train your own model. I would love to do it in ubuntu, because its also free.

@murtazahussain8224

3 жыл бұрын

@@FedericoTerzi fed can u help me with my project .. willing to pay or hire any developer if u can help

Why is teh function of the scorer?

hello i am a visually impaired person how can i get setup file of mozilla tts for screen reader called NVDA.. is it possible

@FedericoTerzi

3 жыл бұрын

Hey, unfortunately, I don't think the deepspeech project is good enough yet for your needs...

Thanks. I can't find the follow-up video though

@DaeOh

Жыл бұрын

Nevermind, I used Whisper for this application!

Thank you so much, but if anyone has done this in a python file kindly please share the link. I'm facing some problems kindly please if anyone has done please provide the link. I have a deadline coming up, please help me.

Its Really cool! I tried this its working but its not printing the text which got listened. Is anyone here faced this? Please help me to fix this.

What about another language? does it generate text too?

@FedericoTerzi

3 жыл бұрын

Yes, it can also do other languages, but you have to find the right model. A good starting point is googling "deep speech spanish model" (replacing spanish with your language of choice) :)

hello this computer operates on a linu system,right?

@FedericoTerzi

3 жыл бұрын

Yes, tt also works on linux :)

1) Python 3.6.5 doesn't work. I updated to 3.6.7 2) activate give an error ... edit activate.bat in Scripts folder and put and '.' after "delims=:" in line 4 then execute Scripts\activate.bat explicitly

Hey, thank you for tutorial. But i have error "Could not find a version that satisfies the requirement deepspeech (from versions: none)" when i enter "pip install deepspeech" command. What i'm doing wrong? I followed your tutor step by step

@FedericoTerzi

3 жыл бұрын

Hey, are you sure you are using Python 3.6?

@pecka6612

3 жыл бұрын

@@FedericoTerzi No, but in latest release (0.8.2) indicates support Python 3.5, 3.6, 3.7 and 3.8, and i'm trying to use 3.7.4. So i think its should work, but i'll gona try with 3.6

@pecka6612

3 жыл бұрын

@@FedericoTerzi Yeah, its working with 3.6. Thank you!!

@FedericoTerzi

3 жыл бұрын

Glad to hear that!

@captaincommand-devminecraf2210

3 жыл бұрын

@@FedericoTerzi me im using python 3.6.7 like you but have the same error :c (im french)

Hi, can you make a tutorial on how to train your own model for deepspeech

@FedericoTerzi

4 жыл бұрын

Yes, that could be the topic of a future video :)

@piotrszostakiewicz7925

3 жыл бұрын

​@@FedericoTerzi can't wait to watch this!

@chetanmundhe8619

3 жыл бұрын

Yes, please make video on how to train deepspeech model