---
title: README
emoji: 🌬️
colorFrom: blue
colorTo: indigo
sdk: static
pinned: false
---

# Windstorm Labs

[Windstorm Labs](https://windstormlabs.com) is part of the [Windstorm Institute](https://windstorminstitute.org).

**For developers, the canonical organisation is [WindyTranslate](https://huggingface.co/WindyTranslate)**, with scores, licences and attribution for every model at **[windytranslate.com](https://windytranslate.com)**.

## What is here

- 1,535 unmodified mirrors of upstream models (`origin-*`), mostly [OPUS-MT](https://huggingface.co/Helsinki-NLP) and [HPLT](https://huggingface.co/HPLT). Their model cards are the upstream cards, unchanged, with the upstream licence text added.
- 1,612 translation models (`translate-*`), copies of the models also published in WindyWord. Where a canonical copy exists in WindyTranslate, the card links to it.
- 66 speech-recognition models (`listen-windy-*`, `windy-*-ggml`), derived from OpenAI Whisper, Distil-Whisper and community Whisper fine-tunes. Each card names its upstream.
- 7 video-generation mirrors (`wan22-*`, `ltx2-19b-fp8`, `hunyuan`) used by SceneMachine, with the upstream licence files.

## Scores

Translation screening scores (chrF++ and BLEU on 48 FLORES-200 dev sentences) are published on [windytranslate.com](https://windytranslate.com) and on the WindyTranslate cards. They are a first filter, not a publication result.

## Licensing

Every model keeps its upstream licence (Apache-2.0, CC-BY-4.0 or MIT; the video mirrors keep their vendors' community licences). Each repository carries the licence text and an attribution notice. Two groups are marked **licence under review** and are not for commercial use until the notice is removed: the WindyScripture variants (trained on the eBible corpus, parts of which are non-commercial / no-derivatives) and five speech models whose upstream declares no licence.