Bluzen Labs is a passion project run as a non-profit community. We build the models, datasets and demos this corner of the field keeps needing and nobody funds, then give all of it away. Fandom deserves infrastructure as solid as anything else in machine learning.
Weights, corpora and demos, published as we finish them. These are the most-used entries in each collection.
Weights we have trained and published, with the recipe attached.
Corpora, annotations and benchmarks, released with their filters.
Demos you can open in a browser and break in about a minute.
Four of the twelve. The rest, including the ones that did not work, are on the projects page.
The dataset everything else is trained on
An 18M-image illustration corpus with tags, captions and quality bands, plus the whole filtering pipeline that produced it.
Multi-label tagging for illustration
A tagger over a 12k-tag vocabulary that holds up on the long tail, where general-purpose vision models quietly collapse into the fifty most common labels.
Adapter training that fits on one GPU
The trainer the rest of the lab runs on. Adapter families, quantised paths and a config format you can read out loud.
Flow-matching generation with region control
A text-to-image stack built around region conditioning, so a prompt can say where something belongs rather than only what it is.
Method write-ups, post-mortems and the occasional argument about evaluation. Nothing here is a press release.
Why we throw away aggregate F1 on tagging benchmarks, and what we report instead. A short argument with a long table attached.
Inpainting treats a mask as damage. PrismStream treats it as composition. Notes from six weeks of getting that distinction to survive training.
On building Aurora: why every quality filter in the pipeline emits a column instead of deleting a row, and what that costs in storage.
Kubernetes answers a question small labs are not asking. Notes on writing a scheduler whose main feature is explaining itself.
We do not read CVs. Fix something small, stick around, and you are a member. That is the whole process, and it has not failed us yet.