Telegram Meme Bot

🎥Demo Video🎥

Project

Currently, when you are chatting with your friends and want to send a meme you have to leave the messenger, go to an ad-filled app or website, generate the meme, download it and then send it into the chat. But not anymore! This bot lets you generate meme right inside of Telegram!

Simply pick one of the shuffled images, enter your desired texts and forward the generated image!

Features

• Shuffle: Creates an image where 3 random meme templates are resized to fit on one big image without any black space and guide text is added.
• Select: Pick one of the shuffled images, enter your texts and get the finished meme
• Image arrangement: Based on the width and height of the 3 images, they will either be stitched together horizontally or vertically

Dev Tools

• Docker: To easily deploy and (in the future) scale the app, it is containerized using Docker and can be started with docker compose up
• pre-commit: To ensure consistency in the codebase, pre-commit is used with: flake8, black, mypy
• GitHub Actions: To test the code at each commit, a GitHub actions pipeline was created that executes pre-commit and runs all the tests

Database

For each meme template a document exists in the mongoDb collection. Since no relationships exist in these objects, I decided to use a NoSQL database to avoid the overhead of using an SQL database such as Postgresql.

For each meme template, the x and y coordinates and the width and height of each textbox where text can be entered is saved. Additionally, the name of the template is saved, so that it can be accessed. In the future, if the images where not saved locally but on a service such as Amazon S3, the only thing that would need to be changed is the template-location.

Example document

{
  "id": "9",
  "name": "The Rock car meme",
  "text-locations": [
    {
      "x": 240,
      "y": 180,
      "width": 155,
      "height": 212
    },
    {
      "x": 414,
      "y": 183,
      "width": 160,
      "height": 211
    }
  ],
  "template-location": "./meme9.jpeg"
}

Future Improvements

To make the bot more usable and scalable, some of the features listed here could be implemented:

Let a user select a language

Currently to change the language, the settings variable has to be changed
and the server restarted. A feature could be added that lets users
pick the language in which the bot communicates with them

Allow users to search for specific meme templates

Currently the user has to call /shuffle until he finds a meme
template that fits his needs. To avoid this, a feature could be
added to let users search for specific templates

Separate telegram interaction and image generator into individual microservices

The part of the code that generates the code could be moved out into a separate
docker container, to be able to scale it independetly without creating more
instances of the container that interacts with the Telegram API

More Tests

Tests for test ImageShuffler class have already been added.
But of course more tests for the other classes could be added to increase
test test coverage of the project

Cloud Blob Storage

Currently all the templates are saved on the host server,
which adds additional workload on a single machine.
To prevent this and use more storage, a cloud service such as
AWS S3 Cloud Storage could be used to store the images in the cloud.

Add more meme templates

Currently there are only 10 meme templates in the database.
More templates should be added to give the user more options to choose from

Outro

Thank you for making it this far and for taking my project into consideration. If you have any more questions, don't hesitate to write me either via Slack or email: hoppe.florian02@gmail.com

Also if you want to learn more about me, feel free to check out my website

💻Personal Website💻