This project was an ambitious hobby experiment that blended together Instagram scraping, AI-driven video commentary, and creative overlays.
The core idea was to grab funny or trending clips from meme accounts automatically, then process them through a pipeline that gave the illusion of two cats, Sunny and Dusty, reacting to the videos in their own quirky personalities.
- Uses Instaloader and Instagrapi to log into an Instagram account (via
credentials.txt). - Scrapes the latest videos from target meme pages.
- Uploads finished reaction videos back with generated captions.
- Relevant files:
instascraper.py,uploader.py.
- Extracts frames and analyzes them with a Hugging Face image captioning model + OCR (Tesseract).
- Generates witty scripts and captions in the voices of Sunny and Dusty using OpenAI GPT models.
- Relevant file:
brianrot.py.
- Dialogue voiced through Amazon Polly (with separate mappings for each cat).
- Extra cat sound effects layered at specific timestamps.
- Relevant files:
brianrottts.py,catsoundoverlay.py.
- Cartoon cat avatars and dynamic subtitles placed in video corners.
- Synchronization with script lines and emotional tags.
- Relevant files:
emotionoverlay.py,subtitlesoverlay.py.
- Tools like ImageMagick, Tesseract, and MoviePy stitched everything into a polished “reaction” clip.
At its peak, the pipeline could fully automate a reaction-style video:
from scraping a clip → generating dialogue → producing a finished, captioned, voiced, and watermarked output ready for upload.
Now, however, the project is in a broken, archival state.
Several dependencies have changed or drifted out of sync (OpenAI model calls, AWS Polly authentication, Instagram API behavior).
Hard-coded paths, outdated packages, and fragile chaining mean the workflow no longer runs end-to-end.
What remains is a fascinating snapshot of a creative coding experiment — equal parts machine learning demo, video editing hack, and playful internet automation — that served as a passion project rather than a production tool.