README.md

Captionaize 🤖

A generative AI tool that uses Google Gemini 1.5 to parse through your videos — suggesting captions and hashtags that are tailored to your content

The tool suggest different captions/hanstags for TikTok or Instagram, since trends vary on either platform

How to Use - App

There is a nice UI that you can use to interact with the tool here

Upload your video file and away you go

How to Use - Code

Alternatively, the app can be run by cloning this repo and working with it locally

• Clone repo

  git clone https://github.com/tjoab/captionaize.git
  cd captionaize

• Set up your virtual enviroment using Python 3.11

  python3 -m venv venv
  source venv/bin/activate

• Install dependencies pip3 install -r requirements.txt

  (Streamlit dependency not needed if you don't plan on running the streamlit app locally - can be removed)

• Create a new main script file

• Import the helper.py module to your main script

• Add the following to your main script

  from helper import authenticateAPI, uploadVideo, modelInference, deleteVideo
  
  authenticateAPI(YOUR_GOOGLE_API_KEY)
  
  filePath = "path_to_your_video_file_in_project_dir"
  
  videoFile = uploadVideo(filePath)
  
  response = modelInference(videoFile)
  
  deleteVideo(videoFile)

• The response variable contains the raw response from the LLM request, which can be printed directly and inspected, but it is a bit ugly. You could also import the parseResponse() function from the helper module and get a formatted tuple:

  Tuple[str, List[str], List[str]] = (Caption, List of Viral-esque Hashtags, List of Relevance-esque Hastags)

  from helper import parseResponse
  
  platform = "tiktok"
  # or
  platform = "instagram"
  
  parseResponse(response, platform)

🛠️ Built With

• Google Gemini 1.5 Flash - The LLM used
• Streamlit - The UI library used - hosting also done on their servers

📄 License

This project is licensed under the MIT License - see the LICENSE.md file for details

☕ Support

If you find this project helpful, you can support me by buying a coffee: