video-edsun123 created by GitHub Classroom

Main Exercise: Using the twitter feed, construct a daily video summarizing a twitter handle day

1. Convert text into an image in a frame
2. Do a sequence of all texts and images in chronological order.
3. Display each video frame for 3 seconds

This API retrieves tweets from the last 24 hours of tweets from that user's timeline then returns a video of all tweets in that timeline.

Each text from the tweet is converted to text on a frame which is shown for 3 seconds. Then each is saved as a 3 second video in the current directory and a text_file keeps track of those video filenames. The text file is necessary to concatenate all the files into a mega video called mergedfile.mp4. Note: before each call, the program will clean out the current directory and the text file to prepare for writing more mp4 files in the next call.

For the queueing system, I have 1 consumer process that retrieves the tweets and 2 producer processes that convert those tweets to videos. Using the module queue helps lock the processes so that current tasks are finished and also facilitates adding and removing tasks from the queue task system.

1. Main program is run in queue_test.py To run...

python3 tweet_vid_converter.py

2. Video file names are stored in list.txt
3. Keys for authentication is stored in separate keys file called keys.py

TWITTER_API_KEY = XXXXXXXXX
TWITTER_API_SECRET_KEY = XXXXXXXXXXXXXXXXXX
TWITTER_ACCESS_TOKEN = XXXXXXXXX
TWITTER_ACCESS_TOKEN_SECRET = XXXXXXXXXXXXXXXXXXXXXXXXXXX

*Note: if you want to change specific number of tweets and change the tweet handle. Change the variables username and num_tweets in lines 59 and 60. 

Normally you would expect on thread is assigned to each core. Given a mac computer with 4 cores, it should handle 4 threads of tweet-video conversion. However, I've found that the performance jumps around 30-40% in cpu usage instead of 25%. Then it is found that by default, each thread for frame conversion is ~1.5 threads per core. So for performance reasons, I've had at most 2 producers or 2 threads to do the video-thread conversion

#Flask application I have issues conecting to the amazon ec2 instance because ports for 80 is closed. However, I was able to fix the issue by going to an ubuntu server hosted by amazon. I was able to ssh into the webserver, scp files into the server, downloaded the dependencies in the server, and ran it using rest service. There is a folder called eflask2 that helps install dependencies and run the application in the ubuntu web server.