EVOKE-Turbo world model adapted from Alaya-EVOKE:
https://github.com/AlayaLab/Evoke (revision 1d271be45128bbda4263e9e3c7dfc8f6ba260492)
https://huggingface.co/AlayaLab/Evoke-Turbo
Copyright 2025 The Helios Team and The HuggingFace Team, Copyright 2026 The Evoke Team.
The inference code and the Evoke-Turbo weights are licensed under Apache-2.0 (LICENSE in this folder).
The WanGP checkpoint stores the transformer in BF16 with the modules the source keeps in FP32 at inference
(time embedder, modulation tables, the affine cross-attention norm) in FP32, under WanGP's Wan module names.

The world memory's depth model is ViGeo1.1:
https://github.com/aigc3d/ViGeo (code, commit 78100ce), Apache-2.0
https://huggingface.co/pkqbajng/ViGeo1.1 (weights)
Its DINOv2 layers are Copyright (c) Meta Platforms, Inc. and affiliates, Apache-2.0.

THE ViGeo1.1 WEIGHTS ARE LICENSED UNDER CC-BY-NC-4.0 (Creative Commons Attribution-NonCommercial 4.0,
https://creativecommons.org/licenses/by-nc/4.0/): they may not be used for commercial purposes. EVOKE worlds
need them for their memory, so generating or exploring an EVOKE world is subject to this restriction; users
are responsible for meeting it. The WanGP file keeps the trunk and decoder projections in FP16 and the rest in
FP32, without the heads EVOKE does not read (ray, normal, mask, confidence, camera).

The Wan2.1 VAE and the UMT5-XXL text encoder are WanGP's existing Wan2.1 components (Apache-2.0).
