Hy-Nomas world model adapted from Tencent Hy-Nomas:
https://github.com/Tencent-Hunyuan/Hy-Nomas (read at revision 18974d41ada7af4aa3a026480197ab0ac0702682)
https://huggingface.co/tencent/Hy-Nomas
Copyright (C) 2026 Tencent. All rights reserved.
The Hy-Nomas inference code, DiT weights and LoRAs are licensed under Apache-2.0; LICENSE in this folder is Tencent's license file,
with the notices of the third-party components of its distribution. The prompt rewriter instructions in prompts.py are condensed from
the release's online prompt rewriter (Apache-2.0).

The default video decoder is PixelUMM, as fine-tuned for Hy-Nomas (tencent/Hy-Nomas VAE/PixelUMM_FT):
https://github.com/nv-tlabs/PixelUMM (code Apache-2.0; file-specific notices retained upstream)
https://huggingface.co/nvidia/PixelUMM (weights)
THE PIXELUMM WEIGHTS AND THEIR DERIVATIVES ARE LICENSED UNDER THE NVIDIA ONE-WAY NONCOMMERCIAL LICENSE (LICENSE_PixelUMM in this folder):
they may be used for non-commercial research or evaluation purposes only. Hy-Nomas worlds are decoded with them, so generating or
exploring a Hy-Nomas world is subject to this restriction; users are responsible for meeting it. The WanGP file
(hy_nomas/pixelumm_decoder_fp8.safetensors) holds the decoder tensors of the release's FP8 file (VAE/PixelUMM_FT_FP8) unchanged, without
its encoder.

The Qwen2.5-VL-7B-Instruct text encoder is WanGP's existing Qwen-Image component (Apache-2.0).
