Why lore, though?

A short essay for people still keeping textures in git and calling it a workflow.

Somewhere right now, someone on your team is running git clone on a 60GB project and considering a career change. Nobody designed git to store binary assets well. It was built to diff Perl scripts, and it still shows.

The usual fixes don't fix it. Git LFS swaps your files for pointer files and needs a second server you now have to keep alive too. Perforce works fine, the way a toll booth works: it gets you there, for a per-seat fee, forever, with a central server someone has to babysit for the rest of their career.

lore is Epic Games' answer, and Epic has been moving 60GB Unreal projects around for two decades, so they'd know. It started inside Epic as Unreal Revision Control, backing UEFN, before Epic open-sourced it under MIT.

Content is stored by hash, in chunks, deduplicated at the fragment level. A multi-gigabyte texture gets the same treatment as a kilobyte of text, not an afterthought bolted on in a point release. Workspaces hydrate lazily, so you're not pulling the whole asset library down just to open the project, and staging, committing, branching, and diffing never leave your machine to hit a network round trip.

Credit where it's due

Every revision's hash chains to its parents and its data, which makes history tamper-evident by construction: if the chain says nothing changed, nothing changed, and there's no "checksum mismatch, good luck" three years into a project. Branches are lightweight mutable references rather than copies, so you can create, switch, and abandon as many as you like without duplicating a 40GB asset library every time someone wants to try something. The CLI isn't a UI hiding a black box, either: the same surface ships as an API in C/C++, Rust, C#, Go, Python, and JavaScript, so anything you can do by hand you can also script.

It also tells you what it isn't yet, which is rarer than it should be. File locking today informs rather than blocks: it flags that someone's editing a binary, it doesn't stop a second person doing the same thing. lore says so out loud, instead of shipping half a feature as a whole one.

The catch

It's open source, MIT-licensed, no seat fees, no lock-in. It's also pre-1.0, so yes, there are gaps, and the capability reference lists them without flinching. Same honesty as the locking bit: nobody's pretending this is finished, just that it's already better than what you're using now.

Convinced, or just tired of the alternative? Either way, getting started takes about five minutes.