Skip to main content
Back to Blog

Introducing Kohaku Lab

3 min read
KohakuBlueLeaf
#announcement #about
Introducing Kohaku Lab

Welcome to Kohaku Lab

Hey everyone! Welcome to Kohaku Lab - a virtual laboratory made by AI/ML enthusiasts who build tools we actually want to use.

What Is Kohaku Lab?

We’re not your typical research lab. There’s no physical building, no corporate backing, and no formal hierarchy. Just a group of people who got frustrated with existing tools and decided to build better ones.

How It Started

Simple: we all had itches to scratch.

We all use text-to-image models. Commercial APIs are expensive, existing open models had limitations. So we trained HDM at home on consumer hardware.

We all used HuggingFace for model hosting. Then their policies shifted in ways we didn’t like. So we built KohakuHub.

We all used Weights & Biases for experiment tracking. Limitations and pricing got annoying. So we created KohakuBoard.

We needed better fine-tuning techniques. So we developed LyCORIS.

Pattern: Need a tool? Build it yourself. Share it openly. Use it daily.

That’s Kohaku Lab.

The Philosophy

We build things we want to use. That’s the whole philosophy.

If we’re not using it ourselves, why would we build it? If it doesn’t solve a real problem we have, what’s the point?

This means:

  1. We’re our own users - We know if something sucks because we use it daily
  2. Real problems only - No building for the sake of building
  3. Practical over theoretical - If it doesn’t work in practice, it doesn’t work
  4. Share openly - If it helps us, it might help others too

No grand mission statements, no five-year plans. Just people solving their own problems and sharing the solutions.

Open Access

Everything we create is openly accessible. Some projects use permissive open-source licenses, others may use non-commercial licenses to protect our work while still sharing it freely for research and learning.

The point is: if you want to learn from our code, use our tools, or build on our research, you can. No paywalls, no proprietary secrets.

Join Us

Interested in getting involved? Here’s the thing: we don’t have an application process or formal onboarding. Just:

  1. Check out our projects on GitHub
  2. Find something that interests you
  3. Start contributing, asking questions, or sharing ideas

That’s it. If you’re passionate about AI/ML and want to build cool stuff with like-minded people, you’re already qualified.

What’s Next

We’ll be sharing more about our individual projects, technical deep-dives, research insights, and lessons learned. Stay tuned for posts about:

  • The road of HDM
  • A real “hub”, a full-stack ML platforms
  • And whatever else we find interesting!

Thanks for stopping by. Hope to see you around!

Links: