Considering the growing evolution of generative AIs in 2023, we saw them occupying relevance in the most diverse spheres of competence in the modern world: ChatGPT became popular due to its friendly interface, generating texts in chat format; MidJourney with creative visual representations; etc.

One of the most famous recently created was the image generator called Stable Diffusion, a latent text-to-image diffusion model capable of generating photo-realistic images given any text input. Images created by the pre-trained model can be created under many different circumstances, including video compilation of the images created in sync with an audio as input.

Project aimed at creating an Automatic social media channel/page based on fully AI content for image and audio using the deep learning text-to-image model Stable Diffusion and publishing them as short videos using a local entrypoint to it.

All the code functionallity uses Python using versioning control strategies to guarantee quality and security of the application, such as Github Actions and it's custom toolkit. The videos will be extracted from a personal Google Drive and sent to AWS S3 for storage/backup files, created using Terraform.

• Follow the steps through your terminal using the following logic:

  • Clone the repository to your local machine.

    cd to/your/folder/path
    git clone https://github.com/dark-theme-org/ai-shorts.git
    

  • Create a virtual env and install project dependencies.

    • If you don't have Miniconda installed:

      bash Miniconda3-latest-Linux-x86_64.sh
      

    • Create virtual environment with conda (python-version used was 3.9.12)

      cd path/to/github/ai-shorts
      conda create --name=venv python=3.9.12
      

    • Launch venv

      conda activate venv
      

    • Install packages with specified versions

      pip install -r requirements-dev.txt
      

  • Run pre-commit to set up the hooks

    pre-commit install --config .code_quality/.pre-commit-config.yaml
    

  • Configure ./src folder as PYTHONPATH (bash example)

    export PYTHONPATH="$PWD/src"
    

The file src/main.py contains a python script used as entrypoint for local running. You can run it as a script (e.g. cd src and python main.py --scope=prod). The --scope=prod is the environment to be deployed onto. It's an optional parameter which default value is test. This string can be anything, but only the word prod has a different effect, setting the uploaded video privacy as public. Everything else will share the video as unlisted.

Another way, much easier, is to run the file release.sh that add all the necessary for an execution in fewer lines (e.g. sh release.sh prod). In this case, always remember that you need to activate you virtual env with conda activate venv and them run the .sh file.

• .code_quality <> Analyze your source code's quality and complexity, keeping your project's code simple, readable, and easier;
• .github <> Houses workflows, including pull request templates and entrypoint runner;
• notebooks <> Playground for initial coding session;
• src <> Main directory where the entire major scope of the project will be stored;
• test <> Space to create unittests from the functions and methods created.

• .gitattributes <> Control actions in git, passing specific attributes inside the repository;
• .gitignore <> Intentionally targeting extensions that git should ignore when committing to the project;
• AUTHORS.md <> Credits to collaborators who actively participated in this project;
• CHANGELOG.md <> File to report main changes over versions/releases;
• release.sh <> Script to run all the necessary for local entrypoint;
• requirements-dev.txt <> Development dependencies and their used versions;
• requirements.txt <> Code functionallity dependencies.