Streaming with live captions - azure-media-services

Is it possible to stream a live event with live captions in AMS v3. I've seen examples like this doing it with video files. If it is possible, how?

See the following for details on how to accomplish this: https://learn.microsoft.com/en-us/azure/media-services/latest/live-transcription

We have an AI based Live transcription feature in Preview right now. It is limited to en-US only currently.
For live captions generated outside of our service, only pass-through channels currently support embedded CEA 608/708 - which requires your upstream encoder to support feeding that into the elementary stream. In addition, the older Smooth streaming ingest protocol supported captions being sent for live channels. Although overall encoder support for that path in the market is pretty small these days.
Which encoder are you looking at and what is your overall captioning workflow?

maybe you can try BeMyEars if your OS is macOS. Google search with keyword: BeMyEars

Related

Unity SCORM Suspend_Data issue

I am working on a serious game project on Unity for a client that must be playable on a browser and compatible with SCORM.
I discovered the concept of SCORM and LMS with this project.
I made an internal backup system, with a JSON file, which works very well when I build for PC, my score saves well and when I restart my project, I can load my backup.
Unfortunately when I build in WebGL and publish to the SCORM Cloud website, the save and load information doesn't pass.
I found on the internet that I should use suspend_data functions, I did find the description of its functions, but I can't find how to send or retrieve my score.
I must have a problem in the conversion of JSON in string.
Can you help me with this, or give me a hint.
I thank you in advance.
Sincerely
There might be a couple different issues at play here.
First off, I'm not sure that SCORM Cloud allows for the Unity Player mime type today. That could be verified with a request to support#scorm.com and it may be able to be added if it isn't currently supported.
Second, how are you communicating with the SCORM API? When a SCORM course is launched in the player, there is a javascript API exposed with some basic calls you need to make. This is where you'd set completion or a score (or suspend_data if needed) and it tells the LMS what data to persist for the learner session and makes values available from previous launches if set to resume.
This won't happen automatically, so you need to handle that in your Unity package with either a wrapper, or some javascript calls to the proper context from within the package.
Hopefully all this is helpful for your understanding of SCORM. Feel free to reach out to support#scorm.com since you're using SCORM Cloud and they can help you if you need more information.
Note: I work for Rustici Software on the SCORM Cloud team and have some light Unity experience.

Best approach to convert existing video files into mpeg-dash

We have over 50MM videos with an average of 3 different resolution e.g 240, 360 etc.
It is time for us to move to Dynamic Adaptive Streaming or Mpeg-Dash. At the moment our biggest challenge is convert existing data into Mpeg-DASH. Our current approach is to convert all videos one by one and create MPD file, this could take months.
Is there an alternate approach? I am aware of streaming existing files realtime using different tools but this will require Huge CPU resource? Any benchmarks that can help us decide how should we be moving for Mpeg Dash?
Thanks
If you want to convert a huge amount of videos in a short amount of time it might be a good approach to use one of the cloud encoding services available as they can convert multiple files in parallel.
I personally have good experience with bitmovin cloud encoding as I really like their API clients with lots of examples: https://bitmovin.com/encoding/
However, there are other services available as well
encoding.com
Zencoder

create tagging schema

There are ~50k articles which needs to be analyzed and a tag schema dictionary needs to be created post, which becomes the base for the Auto Tagging feature we plan to build for the site.
Are there any tools available to tag the data and help generate a list of kind of data available in the data-set.
If not then what would be a good place to start?
There is a data mining tool called Weka, You can try downloading it. There are many weka videos in youtube. You can also google it there is lot of information available.
Well there are few tools for data analysis and you can watch videos on youtube to learn one of them.
1) Weka, KNIME- if you are good in Java then you should go for it.
2)Orange, NLTK- if you are good in python then you can choose this.
3)R-programming- if you are good in C, then you can go for it.

SMB/samba support on iOS?

I check iOS document and also google it for a while and get the impression that iOS does not support samba (although there is a samba app for jailbreak iPhone).
But then how do the app FileBrowser achieves that? Does that mean they implemented samba support by their own ?
I also find there is an open source library called tango that provides limited support for samba. So my question is that the best samba support I can get ?
I spent quite some time in implementing my own SMB client so I would like to share some experience here.
First do not use tango in your production code because once you become familiar with SMB you will realize that its implementation is problematic, e.g. it does not support unicode and in some several cases it is not correctly padding so you can't access the folder. And I also heard people said they can't connect window 7 with it.
Second, to summarize my experience I find jcifs guys had said the best: "anyone who wants to implement the CIFS needs to know one very important thing - the "official" CIFS documentation is not accurate
and does not reflect reality. There is NO specification. Do not believe anything you read in the IETF draft or the SNIA document (same document different formatting). Use it only as a hint. The definitive reference is whatever you see on the wire.
WireShark Rules!
... look at JCIFS for design inspiration such as how it puts the request and response into a map by MID and encodes and decodes frames.
Then implement the following commands:
SMB_COM_NEGOTIATE
SMB_COM_SESSION_SETUP_ANDX
SMB_COM_TREE_CONNECT_ANDX
SMB_COM_NT_CREATE_ANDX
SMB_COM_READ_ANDX
SMB_COM_WRITE_ANDX
SMB_COM_CLOSE
all responses for above
"
The only thing I can add is that , you also need to implement TRANS2_FIND_FIRST2 request/response to query the files inside a folder and if you want to find out how many shared folders the server exposes you need to implement NetShareEnum Request/Response.
I used libsmbclient from samba package (http://www.samba.org) for SMB operations on iOS.
You can look on my project https://github.com/kolyvan/kxsmb (objective-c wrapper on libsmbclient). For now it supports a limited set of SMB operations. It mostly was designed for browsing local net and retrieving files from SMB shares.
iOS doesn't give you access to a filesystem that you may be used to. You can read and write files inside your own App's private area, but that's all. You could potentially implement another file system in your application, but you won't be able to use normal file operations.
I'd bet FileBrowser implements the protocol inside their app and implements a file system like layer on top of that for access. I'd guess you could either try to port an existing samba library or roll your own.
iOS does not have any APIs to work with SMB. However, SMB is currently documented by Microsoft and implementing it is not impossible (although not easy too). I've created a freeware project which contains most of what you need to handle SMB. See
https://sourceforge.net/projects/smb4ios/

transcribe a phone recording

There is a certain organization that periodically provides information in the form of a recorded message on a "hotline". Is there any open source solution (or set of components that could be "wired" together) that would allow me to present this information in text form on a web page?
Since it's the really easy part, I'm going to assume you can fetch the audio from the "hotline", i.e. you have direct access to the actual audio samples.
The hard part is transcribing the audio. You can start by having a look at Wikipedia and follow the links from there. One solution you could use would be CMU Sphinx. Google and other related search tools such as Google Scholar are likely to become your close friends :)
While there are a number of voice recognition engines available, their accuracy is far from perfect.