How to read Youtube Videos with OpenCV at specific timestamps and set duration without downloading it?

Question

I have tried using pafy but it plays the video from beginning, I want to run my model on specific parts of the video. If this is possible please guide me how to do it.

Any help is appreciated, thanks in advance :)

Answer 1

It was quite easy actually, I figured it out :)

Here's the code:

import cv2
import pafy

#Ask the user for url input
url = input("Enter Youtube Video URL: ")

#Getting video id from the url string
url_data = urlparse.urlparse(url)
query = urlparse.parse_qs(url_data.query)
id = query["v"][0]
video = 'https://youtu.be/{}'.format(str(id))

#Using the pafy library for youtube videos
urlPafy = pafy.new(video)
videoplay = urlPafy.getbest(preftype="any")

cap = cv2.VideoCapture(videoplay.url)

#Asking the user for video start time and duration in seconds
milliseconds = 1000
start_time = int(input("Enter Start time: "))
end_time = int(input("Enter Length: "))
end_time = start_time + end_time

# Passing the start and end time for CV2
cap.set(cv2.CAP_PROP_POS_MSEC, start_time*milliseconds)

#Will execute till the duration specified by the user
while True and cap.get(cv2.CAP_PROP_POS_MSEC)<=end_time*milliseconds:
        success, img = cap.read()
        cv2.imshow("Image", img)
        cv2.waitKey(1)

Answer 2

update for python 3.11

(changes to url parse)

import cv2
import pafy
import urllib

#Ask the user for url input
url = input("Enter Youtube Video URL: ")

#Getting video id from the url string
url_data = urllib.parse.urlparse(url)
query = urllib.parse.parse_qs(url_data.query)
id = query["v"][0]
video = 'https://youtu.be/{}'.format(str(id))

#Using the pafy library for youtube videos
urlPafy = pafy.new(video)
videoplay = urlPafy.getbest(preftype="any")

cap = cv2.VideoCapture(videoplay.url)

#Asking the user for video start time and duration in seconds
milliseconds = 1000
start_time = int(input("Enter Start time: "))
end_time = int(input("Enter Length: "))
end_time = start_time + end_time

# Passing the start and end time for CV2
cap.set(cv2.CAP_PROP_POS_MSEC, start_time*milliseconds)

#Will execute till the duration specified by the user
while True and cap.get(cv2.CAP_PROP_POS_MSEC)<=end_time*milliseconds:
        success, img = cap.read()
        cv2.imshow("Image", img)
        cv2.waitKey(1)

Also, I had to install the pafy package from git source due to like_count not being reliable

pip install git+https://github.com/mps-youtube/pafy.git

See this issue