Git LFS — Large Files

Git was designed for TEXT files with small, line-based changes. It stores full snapshots and uses delta compression to save space. This works brilliantly for source code — a 10-line change in a 500-line file produces a tiny diff.

Binary files (images, videos, ML models, PDFs, compiled artifacts) are completely different. They don't have "lines" — Git cannot delta-compress them effectively. Every modification stores a COMPLETE new copy in the repository.

A 500MB file modified 10 times = 5GB in your .git directory, even though you only need the latest version. The repository grows with every binary change and never shrinks.

Cloning a repo with years of binary history downloads ALL versions, even old ones you'll never use. A 2.5GB repo takes 30+ minutes to clone on a decent connection.

Git operations become painfully slow: git status scans large files, git push uploads massive objects, git clone downloads everything.

GitHub rejects files larger than 100MB by default. Push a 200MB file and you get an error — your team is blocked.

# The large file problem
# Team adds a 500MB design file to the repo
cp ~/design-v1.psd assets/design.psd
git add assets/design.psd
git commit -m "add design file"
git push origin main

# Designer updates the file 5 times
cp ~/design-v2.psd assets/design.psd && git add . && git commit -m "design v2"
cp ~/design-v3.psd assets/design.psd && git add . && git commit -m "design v3"
cp ~/design-v4.psd assets/design.psd && git add . && git commit -m "design v4"
cp ~/design-v5.psd assets/design.psd && git add . && git commit -m "design v5"

# Now the repo is 2.5GB (5 versions x 500MB)
du -sh .git/
# 2.5G

# New developer clones the repo
git clone https://github.com/company/app.git
# Downloads ALL 2.5GB — even the 4 old versions they don't need!
# Takes 30+ minutes

# GitHub rejects the push:
# remote: error: File assets/design.psd is 500.00 MB;
# this exceeds GitHub's file size limit of 100.00 MB