I wanted to replicate Tom Scott's "This Video Has x Views" video ever since I first saw it. And I was able to. Mostly.

Here is an example: https://www.youtube.com/watch?v=tc_1Aqa2u9k

Functionality

I was able to update the title, the thumbnail, and access certain statistics about the video and channel to then display in the video description.

I was starting to experiment with the localizations/translations by translating to Spanish manually. While this was good practice, I realized I could automate this. I translated the video title and description into every available language using the deep_translator library. So if a user changes the language of YouTube in their browser or is using a language different from English, it will translate the video title and description to that language. I used threading to make this process happen faster which would be especially important if I wanted to get this to run every second.

The thumbnail is updated only once a day UTC time and just displays the time. It was created using the PIL library. There was a request limit specifically for the thumbnail which wasn't specified. See limitations section below.

Deployment Process

I created an AWS EC2 instance, cloned the code repository into the instance, and tested that the code worked in the instance.

I also tested that it worked using the AWS-RunShellScript command document in AWS Systems Manager.

Then I created an AWS Lambda function for starting the same EC2 instance, running the code in the instance, and stopping the instance.

import boto3

region = 'us-east-2'
instances = ['i-xxxxxxxxxxxxxxxxx']

ec2 = boto3.client('ec2', region_name=region)

def lambda_handler(event, context):
    ec2.start_instances(InstanceIds=instances)
    print('started instances: ' + str(instances))

import json
import boto3
import urllib.request

def lambda_handler(event, context):
    region = 'us-east-2'

    ec2_client = boto3.client('ec2', region_name=region)
    ssm_client = boto3.client('ssm', region_name=region)

    client = boto3.client('ssm')

    params = {
        'commands': [
            'python3 logic.py'
        ],
        'workingDirectory': [
            '/home/ec2-user/...'
        ],
        'executionTimeout': [
            '3600'
        ]
    }

    response = ssm_client.send_command(
        DocumentName='AWS-RunShellScript', 
        InstanceIds=['i-xxxxxxxxxxxxxxxxx'],
        Comment='Running the youtube update video script.',
        TimeoutSeconds=600,
        Parameters=params
    )

import boto3

region = 'us-east-2'
instances = ['i-xxxxxxxxxxxxxxxxx']

ec2 = boto3.client('ec2', region_name=region)

def lambda_handler(event, context):
    ec2.stop_instances(InstanceIds=instances)
    print('stopped instances: ' + str(instances))

I then used Amazon Eventbridge schedules to create cron jobs for the three Lambda functions.

Stopping the instance, rather than keeping it running 24/7, helps to keep AWS costs down.

Because of the 10k quota limit in the YouTube API I created a script to figure out which times would work best for the schedules. The optimal interval for running the video update logic came out to 9 minutes. I'm not sure how Tom Scott was able to run essentially every second. It could be something from him requesting a quota increase, which I may do in the future (if I do I will update), or he was grandfathered in to some old API functionality.

All code was done in Python. I would like to, one day, make the code available on GitHub but I have some limitations or issues named below I wanted to try solving first.

Limitations

Quota

I was limited to 10k quota, as I mentioned above.

Documentation

The YouTube API documentation was not the best. It could stand to be improved.

Back end error

The video description I was uploading was sometimes causing an uninformative backend error. Sometimes the description worked and sometimes it didn't. It was a gamble even though the information was mostly the same.

Description update lag

The YouTube studio video description would update faster than the viewer YouTube video.

Sometimes the description for the viewer YouTube video would be in sync but then in would randomly stop being in sync. There seems to be some YouTube back end issue which I've made bug request for and asked about on Stack Overflow just to be redirected to the bug request website.