Pipe downloading video to opencv

[austinmilt]

Please follow the guide below

• You will be asked some questions and requested to provide some information, please read them carefully and answer honestly
• Put an x into all the boxes [ ] relevant to your issue (like that [x])
• Use Preview tab to see how your issue will actually look like

---

Make sure you are using the latest version: run youtube-dl --version and ensure your version is 2016.11.22. If it's not read this FAQ entry and update. Issues with outdated version will be rejected.

• [X ] I've verified and I assure that I'm running youtube-dl 2016.11.22

Before submitting an issue make sure you have:

• [X ] At least skimmed through README and most notably FAQ and BUGS sections
• [ X] Searched the bugtracker for similar issues including closed ones

What is the purpose of your issue?

• Bug report (encountered problems with youtube-dl)
• Site support request (request for adding support for a new site)
• Feature request (request for a new functionality)
• [ X] Question
• Other

---

The following sections concretize particular purposed issues, you can erase any section (the contents between triple ---) not applicable to your issue

Description of your issue, suggested solution and other information

I am writing a program to analyze the contents of a Youtube video using opencv. Ideally, I would like to be able to analyze the video while it is downloading (i.e. as if it were streaming on Youtube), and then throw away frames that have already been analyzed without ever writing to disk. Is it possible for me to pipe youtube_dl's download directly to another process rather than to disk, or some other workaround that would achieve my aims?

[yan12125]

Like this:

youtube-dl -o - <the URL> | your_program

[austinmilt]

Thanks for your reply yan12125, but you didnt answer my question. Perhaps I need to be more explicit...

Let's say I have a Youtube URL and I'm interested in analyzing the video without saving it to disk. Further, I am interested in doing this analysis using opencv in Python. In my Python script, I would write:

import cv2, youtube_dl
options = {'outtmpl': #SOMETHING#}
Y = youtube_dl.YoutubeDL(options)
Y.extract_info(r'http://some.url')

Does this make more sense?

[yan12125]

The youtube_dl Python library doesn't support streaming to Python objects yet, so the current solution is using subprocess.Popen. For YouTube, things are simpler as they serve plain HTTP files, so you can do something like this:

info_dict = Y.extract_info(url, download=False)
video_stream = urllib.request.urlopen(info_dict['url'])
# video_stream is a file-like object with read() method

[austinmilt]

Thanks for your response. I am able to get a stream as you have indicated, but I dont understand what to do with it. How do I decode the stream into a useable format? Others suggest that for live streams you can use opencv's cv2.imdecode and numpy's numpy.fromstring, but I am only getting nonsense from this. Is there further I can do using youtube_dl or do I need to start looking elsewhere after I get the url?

[yan12125]

YouTube videos are served as mp4 or webm files. Those files are usually compressed, so you need to decompress them first to get frame-by-frame images.

[austinmilt]

Thanks for your help. I'm at a loss for now and can live with downloading the video.