Wan 2.1 is worth keeping when an existing local workflow already delivers usable game footage. A release deadline is a poor time to rebuild that pipeline without comparing the results. Its 1.3B text-to-video model offers a smaller 480P starting point; the 14B variants cover text- and image-to-video tasks, with separate VACE weights for editing. Compare matching tasks and resolutions before switching to Wan 2.2. For adult-game character shots, review appearance and motion clip by clip. Keep community adapters and redistributed checkpoints distinct from the official release when checking licenses.
Commercial use & license
The official Wan 2.1 weights use Apache 2.0. Check derivative licenses and the rights to input assets separately.
Adult content rules
The README prohibits sharing unlawful or harmful content, personal data intended to cause harm and misinformation, or targeting vulnerable people. Input rights and any hosting policy still apply; Apache 2.0 is not adult-content approval.
What you need
Use the official dependencies, including PyTorch 2.4.0 or later. T2V 1.3B is recommended at 480P; T2V 14B supports 480P/720P, and I2V 14B uses separate resolution checkpoints. VACE has its own 1.3B/14B weights. The repository's 8.19 GB VRAM figure applies to T2V 1.3B, not the whole family; memory needs depend on the checkpoint and inference settings.
Ranks are recommendations for the task described, based on reviewed documentation. Model ranks cover non-explicit development work. We have not run practical benchmarks. Check content policies and commercial terms separately. How the ranks work →