ReentryMem inference release Wan inference modules derive from Wan2.2 and the BiWM camera-conditioned snapshot. Copyright 2024-2025 The Alibaba Wan Team Authors. All rights reserved. Wan2.2: https://github.com/Wan-Video/Wan2.2 BiWM: https://github.com/LynnReal-AI/BiWM These components are distributed under Apache-2.0; see LICENSE. The overshoot-and-re-noise numerical update follows AMO Sampler: Hu et al., AMO Sampler: Enhancing Text Rendering with Overshooting, CVPR 2025. https://arxiv.org/abs/2411.19415 https://github.com/hxixixh/amo-release ReentryMem supplies calibrated video-entity support and its presence-aware gating. Changes in this inference snapshot: keep camera-text conditioning and checkpoint parameter names; remove training launchers, optimizers, dataset caches, auxiliary losses, intervention experiments and memory-module execution. Add standalone condition-video inference, subject-region sampling and model-release tools.