diff --git a/video_gen_14d/PAPER_PLAN.md b/video_gen_14d/PAPER_PLAN.md
new file mode 100644
index 0000000000000000000000000000000000000000..21bb367ed0fc36fe6cf4cc1f59c9aa05b32c1966
--- /dev/null
+++ b/video_gen_14d/PAPER_PLAN.md
@@ -0,0 +1,686 @@
+# CtrlGuard 论文计划与两周执行汇报
+
+更新时间：2026-07-14  
+项目目录：`/public/interns/hubin/jmtseng/video_gen_14d`
+
+## 0. 给导师/组会的一页汇报
+
+### 我们准备做什么
+
+我们计划研究一个目前开源视频控制模型普遍忽略、但真实使用中必然出现的问题：
+
+> 当 depth、optical flow、pose 等结构控制信号存在局部错误、缺帧、时延、尺度漂移或提取器域偏移时，视频生成模型能否判断哪些控制 token 可信，并只在可信区域服从控制？
+
+我们以开源的 **Wan2.1-VACE-1.3B** 为主基座。原始 VACE 在 15 个控制层中使用同一个标量 `vace_scale` 注入控制 hint：
+
+\[
+x_l \leftarrow x_l + s \cdot h_l
+\]
+
+我们提出 **CtrlGuard**，把它改为按 layer、diffusion timestep、time-space token 自适应的可靠性门控：
+
+\[
+x_l \leftarrow x_l + s \cdot g_{l,t,p} \odot h_l,\qquad g_{l,t,p}\in[0,1]
+\]
+
+门控器只包含约 0.5–3M 参数。它根据控制特征、解析或估计的置信度、局部时序一致性、扩散 timestep 和 layer id，降低错误控制的注入强度；基座模型冻结，必要时再增加 rank-16 LoRA。
+
+### 为什么值得做
+
+1. VACE、ControlNet 类工作默认控制干净，并没有系统研究结构条件污染。
+2. C³ 预测的是生成结果的不确定性，不是外部控制信号是否可信。
+3. ConfCtrl 只针对双视图相机插值和点云投影误差，不是通用 depth/flow/pose 控制。
+4. CAT-Video 主要向文本或 latent conditioning 注入训练噪声，并没有研究 VACE 控制 hint 的 token 级可靠性。
+5. 该问题容易定量、容易做可视化，同时不需要基础模型预训练。
+
+### 计划形成的三项贡献
+
+1. **CtrlBench-C**：首个面向统一视频控制模型的结构条件污染 benchmark，覆盖 synthetic corruption、真实 extractor shift 和局部冲突。
+2. **CtrlGuard**：按 token、layer、timestep 调节 VACE hint 的轻量可靠性门控。
+3. **系统结论**：回答不同控制模态、扩散阶段和网络层对不同污染的敏感性，并给出质量—控制—鲁棒性的 Pareto 分析。
+
+### 两周内成功的定义
+
+- 原始 VACE 在中等污染下，至少两类控制指标下降 ≥15%，且下降显著大于 seed 方差。
+- CtrlGuard 恢复 ≥30% 的污染损失。
+- 干净条件性能下降 ≤2%。
+- CtrlGuard 显著优于 corruption augmentation + 普通 LoRA。
+- 在未见 corruption family 或未见 extractor 上仍有效。
+
+若这些条件未满足，不继续消耗算力包装算法论文。
+
+---
+
+## 1. 暂定标题
+
+### 英文标题
+
+**CtrlGuard: Reliability-Calibrated Structural Control for Video Diffusion under Corrupted Conditions**
+
+备选：
+
+- **When Video Controls Lie: Token-wise Reliability Gating for Robust Controllable Generation**
+- **CtrlBench-C and CtrlGuard: Benchmarking and Mitigating Corrupted Conditions in Video Generation**
+
+### 中文标题
+
+**CtrlGuard：污染结构条件下的可靠性校准视频控制**
+
+---
+
+## 2. 论文核心论点
+
+论文只围绕下面一个核心论点展开，不扩成泛化的视频生成大课题：
+
+> 固定强度注入结构控制会把提取错误逐层写入视频 latent；显式估计控制可靠性并在 VACE hint 层做 token-wise gating，能够在不损害干净控制能力的情况下显著提高污染条件鲁棒性。
+
+为支撑这个论点，需要依次证明：
+
+1. 控制污染确实会稳定、显著地破坏 VACE，而不是随机 seed 造成的波动。
+2. 破坏主要来自错误 hint 被持续注入，而不是单纯生成质量差。
+3. 全局调低 `vace_scale` 无法同时保留正确区域和拒绝错误区域。
+4. token-wise、layer-aware、timestep-aware 的 gate 优于全局 scale、mask 和普通 augmentation。
+5. 模型能泛化到未在训练中出现的污染与提取器。
+
+---
+
+## 3. Abstract 草稿
+
+Controllable video diffusion models increasingly rely on structural conditions such as depth, optical flow, and human pose. Existing methods, however, implicitly assume that these conditions are accurate and temporally aligned, while real-world control signals produced by perception systems frequently contain local failures, missing observations, temporal lag, and domain-dependent bias. We first introduce CtrlBench-C, a benchmark that measures controllable video generation under synthetic corruptions, real extractor shifts, and cross-condition conflicts. Our study reveals that fixed-strength condition injection causes severe and highly localized degradation even when most control tokens remain valid. We then propose CtrlGuard, a lightweight reliability-calibrated gating module that modulates VACE control hints at each network layer, diffusion timestep, and spatiotemporal token. CtrlGuard is trained with clean-condition teacher consistency, corruption supervision, and a clean-preservation objective while keeping the base video generator frozen. Experiments on Wan2.1-VACE across depth, flow, and pose controls show that CtrlGuard recovers control adherence under corruption without sacrificing clean video quality, and generalizes to unseen corruption families and control extractors.
+
+Abstract 中的结果数字必须等实验完成后再填，不能预先承诺。
+
+---
+
+## 4. 与最近工作的边界
+
+| 工作 | 它解决什么 | 与我们的本质区别 |
+|---|---|---|
+| VACE, ICCV 2025 | 统一 generation/editing 与多种控制 | 默认输入控制可靠；固定 `vace_scale` |
+| C³, 2025/2026 | 预测 world model 输出的不确定性 | 估计输出是否可信，而非判断外部控制是否可信 |
+| ConfCtrl, 2026 | 双图像大视角相机插值 | 只针对点云投影与 camera pose，不覆盖通用结构控制 |
+| CAT-Video, 2025/2026 | corruption-aware latent/text conditioning | 不研究结构控制提取器错误和 VACE hint 注入 |
+| ControlNet++ | 强化图像生成的控制一致性 | 图像方法；不处理视频中的时序污染和延迟 |
+| Ctrl-Adapter / Tea-Adapter | 多条件适配和高效条件生成 | 重点不是 control reliability、corruption benchmark 或校准 |
+
+### 我们不能声称
+
+- 不能声称是“第一个鲁棒视频生成方法”。
+- 不能声称是“第一个不确定性视频模型”。
+- 只能声称：在当前检索范围内，这是第一个系统研究 **统一视频生成中的结构控制污染，并在 VACE hint 注入位置做时空可靠性校准** 的工作。
+
+投稿前需要在 arXiv、OpenReview 和 Google Scholar 再做一次最终检索。
+
+---
+
+## 5. CtrlBench-C：具体 benchmark 设计
+
+### 5.1 主控制模态
+
+两周主实验只锁定三类，不一开始扩到所有 VACE 功能：
+
+1. **Depth**
+2. **Optical flow**
+3. **Human pose**
+
+Mask/layout/scribble 只作为 stretch experiment。三类主模态已经覆盖稠密几何、稠密运动和稀疏关键点三种性质。
+
+### 5.2 合成污染
+
+每种污染设置 5 档强度；训练只看其中部分组合，测试留出完整 corruption family。
+
+#### Depth
+
+- 局部 hole：随机块或物体区域缺深度。
+- scale drift：局部或全局深度尺度乘法漂移。
+- offset bias：深度整体偏移。
+- edge blur：几何边界模糊。
+- temporal lag：深度视频延迟 1/2/4/8 帧。
+- frame hold/drop：重复上一帧或随机缺帧。
+
+#### Optical flow
+
+- magnitude scaling：流幅值缩放。
+- angular bias：流方向旋转。
+- local vector outlier：局部随机错误运动。
+- forward/backward inconsistency。
+- temporal shift。
+- missing moving object：抹去前景运动区域。
+
+#### Pose
+
+- keypoint jitter。
+- limb deletion。
+- left/right swap。
+- confidence drop。
+- temporal freeze。
+- identity swap：多人交叉时交换局部骨架。
+
+### 5.3 真实提取器偏移
+
+训练和测试使用不同提取器，避免只学习合成噪声：
+
+- Depth：Depth Anything V2 ↔ MiDaS。
+- Flow：RAFT ↔ UniMatch 或 GMFlow。
+- Pose：DWPose ↔ OpenPose/RTMPose。
+
+第一阶段使用 VACE 已提供或容易安装的提取器。第二提取器只需生成评测控制，不参与训练。
+
+### 5.4 数据划分
+
+#### 训练
+
+- OpenVid-1M 中抽取 2,000–5,000 个片段。
+- 每段裁为 49 frames、480×832。
+- 离线缓存 clean depth/flow/pose 和 confidence。
+- 每次训练在线随机采样 corruption，不存所有污染版本。
+- 只发布视频索引、控制提取脚本和派生元数据，不重新分发上游视频。
+
+#### 验证
+
+- 200 个与训练 source video 不重合的 OpenVid 片段。
+- 用于选择 loss weight、gate 结构和早停。
+
+#### 最终测试
+
+- VACE-Benchmark 中对应的真实与合成 control 样本。
+- 额外整理 100 个真实视频片段，按 source 隔离。
+- 不使用 VACE-Benchmark 训练最终模型。
+
+### 5.5 评测规模
+
+#### D4 screening
+
+- 50 clips
+- 3 modalities
+- 8 corruption families
+- 3 severity levels
+- 2 seeds
+
+合计约 7,200 个视频，用于确认研究问题是否成立。
+
+#### 最终主表
+
+- 100 clips
+- 3 modalities
+- 6 个筛选后的关键 corruption
+- 5 severity levels
+- 3 seeds
+
+合计约 27,000 个视频。生成任务按 GPU 独立并行。
+
+---
+
+## 6. CtrlGuard 方法细节
+
+### 6.1 VACE 中的修改位置
+
+DiffSynth 当前实现位于：
+
+`third_party/DiffSynth-Studio/diffsynth/pipelines/wan_video.py`
+
+原始逻辑：
+
+```python
+current_vace_hint = vace_hints[vace.vace_layers_mapping[block_id]]
+x = x + current_vace_hint * vace_scale
+```
+
+改为：
+
+```python
+gate = reliability_gate(
+    hint=current_vace_hint,
+    reliability_tokens=reliability_tokens,
+    timestep=timestep,
+    layer_id=block_id,
+)
+x = x + current_vace_hint * vace_scale * gate
+```
+
+当 `gate=1` 时，数值上必须复现原模型；这是第一个实现单元测试。
+
+### 6.2 Gate 输入
+
+对每个 VACE hint token 使用：
+
+- 对应位置的 control latent。
+- 解析 confidence map。
+- 局部 temporal residual。
+- hint 的 norm / entropy 等轻量统计。
+- diffusion timestep embedding。
+- VACE layer embedding。
+
+不直接把完整主干 hidden state 输入 gate，避免 gate 变成另一个大模型。
+
+### 6.3 Gate 结构
+
+初版结构：
+
+1. 3D Conv reliability encoder：`C → 32 → 64`。
+2. 对齐到 VACE token grid。
+3. timestep MLP：128 dims。
+4. layer embedding：32 dims。
+5. 两层 MLP：`(64+128+32+hint_stats) → 128 → 1`。
+6. sigmoid 输出标量 gate `[B, N, 1]`。
+
+预计参数量 0.5–3M，小于 1.3B 基座的 0.25%。
+
+### 6.4 Reliability target
+
+对合成 corruption 可以直接得到 token 级 target：
+
+\[
+r^*_{t,p} = 1 - \mathrm{normalized\_corruption\_error}_{t,p}
+\]
+
+真实提取器没有 ground truth confidence 时使用代理：
+
+- Flow：forward-backward consistency。
+- Pose：keypoint detection confidence + temporal acceleration。
+- Depth：flow-warped temporal residual + edge consistency。
+
+### 6.5 损失函数
+
+\[
+\mathcal L =
+\mathcal L_{\mathrm{FM}}
++ \lambda_{\mathrm{teacher}}\mathcal L_{\mathrm{teacher}}
++ \lambda_{\mathrm{gate}}\mathcal L_{\mathrm{gate}}
++ \lambda_{\mathrm{clean}}\mathcal L_{\mathrm{clean}}
+\]
+
+具体定义：
+
+1. `L_FM`：DiffSynth 现有 flow-matching SFT loss。
+2. `L_teacher`：同一视频、prompt、seed、timestep 下，污染输入预测靠近 clean control teacher。
+3. `L_gate`：gate 与已知 corruption reliability target 的 BCE/L1。
+4. `L_clean`：clean control 下 gate 接近 1，且预测不偏离原始 VACE。
+
+初始权重：
+
+- `λ_teacher = 1.0`
+- `λ_gate = 0.2`
+- `λ_clean = 0.5`
+
+只做 `{0.5, 1.0, 2.0}` 的小网格，不展开大规模超参搜索。
+
+### 6.6 两阶段训练
+
+#### Stage A：gate-only
+
+- 冻结 Wan DiT、VAE、T5 和 VACE blocks。
+- 训练 400 steps。
+- 目标：快速确认 gate 是否有能力恢复污染区域。
+
+#### Stage B：gate + LoRA
+
+- 保留 gate。
+- 在 VACE `after_proj` 和必要的 q/k/v/o 上加 rank-16 LoRA。
+- 再训练 600 steps。
+- 若 Stage A 已达到目标，Stage B 只作为增强和消融，不改变论文主方法。
+
+### 6.7 初始训练配置
+
+| 配置 | 值 |
+|---|---|
+| Base model | Wan2.1-VACE-1.3B |
+| Resolution | 480×832 |
+| Frames | 49 train / 81 final eval |
+| Precision | BF16 |
+| GPU | 8×H200 140GB |
+| Per-GPU batch | 1 |
+| Gradient accumulation | 1–2 |
+| Gate learning rate | 1e-4 |
+| LoRA learning rate | 5e-5 |
+| Optimizer | AdamW |
+| Weight decay | 0.01 |
+| Gate-only steps | 400 |
+| Gate + LoRA steps | 600 |
+| LoRA rank | 16 |
+| Gradient checkpointing | 开启 |
+| Seed | 3 个训练 seed 只跑最终配置；开发阶段固定 1 个 |
+
+D1 必须用本机实测吞吐和显存替换估算值。
+
+---
+
+## 7. Baselines
+
+所有 baseline 使用同一 prompt、condition、seed 和采样步数。
+
+1. **VACE-Original**：`vace_scale=1.0`。
+2. **Global Scale Oracle**：在 `{0, 0.25, 0.5, 0.75, 1.0}` 中为每个 corruption 选最佳全局 scale。
+3. **Control Dropout**：训练时随机丢弃整段控制。
+4. **Corruption Augmentation LoRA**：看到相同污染，但没有 reliability gate。
+5. **Analytic Confidence Mask**：不训练，直接用解析 confidence 乘 hint。
+6. **Control Denoiser**：小型 3D Conv 先恢复 control，再送入原始 VACE。
+7. **CtrlGuard Gate-only**。
+8. **CtrlGuard Gate + LoRA**。
+
+Global Scale Oracle 很重要：如果它已经达到 CtrlGuard 的效果，token-wise gate 的必要性就不成立。
+
+---
+
+## 8. 指标与统计
+
+### 8.1 控制遵循
+
+- Depth：scale-aligned AbsRel、RMSE、edge accuracy。
+- Flow：EPE、angular error、temporal warping error。
+- Pose：PCK、OKS、missing-joint rate、ID-switch rate。
+
+### 8.2 鲁棒性
+
+- mCE：mean Corruption Error。
+- severity AUC。
+- clean-to-corrupt relative drop。
+- worst-group score。
+- seen → unseen corruption generalization gap。
+- seen → unseen extractor generalization gap。
+
+### 8.3 生成质量
+
+- VBench：subject consistency、background consistency、motion smoothness、dynamic degree。
+- VideoScore2：visual quality、text alignment、physical/common-sense consistency。
+- VideoAlign：VQ/MQ/TA，仅用于快速筛选，不作为唯一最终指标。
+
+### 8.4 校准
+
+- Gate-error AUROC/AUPRC。
+- Expected Calibration Error。
+- reliability diagram。
+
+### 8.5 人评
+
+- 200 对视频。
+- 每对 3 名评审。
+- 问题分成 control adherence、temporal quality、overall preference。
+- 随机左右顺序和方法匿名。
+- 报告 win/tie/loss、95% bootstrap CI 和 Fleiss' κ。
+
+### 8.6 显著性
+
+- 自动指标使用 paired bootstrap 95% CI。
+- 主结论使用 paired permutation test。
+- 多 corruption 比较进行 Holm correction。
+- 不只报告挑选后最好的 seed。
+
+---
+
+## 9. 必须出现的图和表
+
+### Figure 1：问题
+
+干净 depth/flow/pose、局部污染控制、原始 VACE 失败、CtrlGuard 恢复，以及 gate heatmap。
+
+### Figure 2：方法
+
+VACE control branch、reliability encoder、layer/timestep embedding、token gate 和 teacher consistency。
+
+### Figure 3：退化曲线
+
+x 轴 severity，y 轴 control adherence；比较 Original、Global Scale、Aug-LoRA、CtrlGuard。
+
+### Figure 4：层与 timestep 分析
+
+展示不同层/噪声阶段学到的 gate 强度，形成方法之外的分析贡献。
+
+### Table 1：CtrlBench-C 主结果
+
+三种模态、主要 corruption 的 mCE 和 clean performance。
+
+### Table 2：OOD 泛化
+
+未见 corruption family、未见 extractor、真实控制错误。
+
+### Table 3：消融
+
+去掉 layer embedding、timestep、confidence、teacher、clean loss；gate-only vs gate+LoRA。
+
+### Table 4：效率
+
+可训练参数、训练 GPU-hours、推理显存、延迟和视频质量。
+
+---
+
+## 10. 两周逐日计划
+
+### D1：工程跑通
+
+- 完成隔离环境。
+- 单卡生成 17-frame/8-step smoke sample。
+- 8 卡各生成一个 49-frame/20-step 样本。
+- 记录模型加载时间、采样时间、峰值显存。
+
+交付：
+
+- `outputs/baseline/single_smoke_*`
+- `outputs/baseline/eight_gpu_baseline_*`
+- 八卡日志和 manifest。
+
+### D2：评测与数据协议
+
+- 固定训练/验证/测试 source split。
+- 跑通 depth/flow/pose 三类控制指标。
+- 实现 same prompt/seed/timestep 配对。
+
+交付：50 个 clean 样本的指标与人工 spot-check。
+
+### D3：污染生成器
+
+- 每种主模态完成至少 6 个 corruption。
+- 每个 corruption 5 档 severity。
+- 输出可视化 contact sheet 和单元测试。
+
+交付：`corruptions.json` 与样例。
+
+### D4：研究问题验证
+
+- 完成 7,200-video screening。
+- 画 degradation curves。
+- 比较污染效应和跨 seed 方差。
+
+硬门槛：
+
+- 至少两种模态在中等污染下降 ≥15%。
+- 若不满足，停止 CtrlGuard，切换 GenLeak。
+
+### D5：真实 OOD
+
+- 加入第二 depth/flow/pose extractor。
+- 测试 synthetic → real extractor shift。
+- 锁定最终六种 corruption。
+
+### D6：Gate 实现
+
+- 新建 `ReliabilityGate`。
+- 接入 DiffSynth VACE hint。
+- `gate=1` 数值复现测试。
+- 支持保存 gate heatmap。
+
+### D7：Stage A
+
+- 400 steps gate-only。
+- 比较 analytic mask、global scale、gate。
+
+硬门槛：
+
+- 至少恢复 20% 污染损失。
+- clean 下降 ≤2%。
+
+### D8：Stage B
+
+- rank-16 VACE LoRA。
+- 加入 teacher consistency。
+- 比较 augmentation-only LoRA。
+
+硬门槛：
+
+- 最终恢复 ≥30%。
+- 明显优于 augmentation-only。
+
+### D9：消融
+
+- layer/timestep/reliability/teacher/clean loss。
+- 只保留能解释主结论的消融。
+- 当天结束后冻结方法。
+
+### D10：主表
+
+- 100 clips × 3 modalities × 6 corruptions × 5 severity × 3 seeds。
+- 自动汇总 mCE、AUC、worst-group。
+
+### D11：独立质量评测
+
+- VBench、VideoScore2、VideoAlign。
+- 证明不是靠忽略控制换取表面质量。
+- 汇总失败类型。
+
+### D12：人评和 14B stretch
+
+- 启动 200 对盲评。
+- 若主线已达标，在 VACE-14B 上做 corruption transfer 或小规模 adapter。
+- 14B 不是阻塞项。
+
+### D13：统计、图表、复现
+
+- 95% CI、显著性检验。
+- Figure 1–4、Table 1–4。
+- 一键复算主表和环境锁定。
+
+### D14：写作
+
+- 完成论文初稿、supplement 和 README。
+- 逐条核对 abstract 中每个 claim。
+- 做最终 Go/No-Go。
+
+---
+
+## 11. 算力与存储预算
+
+理论最大预算：
+
+\[
+8 \times 24 \times 14 = 2688\ \text{H200 GPU-hours}
+\]
+
+项目硬上限设为 1,100 GPU-hours：
+
+| 阶段 | GPU-hours 上限 |
+|---|---:|
+| 环境、smoke、吞吐标定 | 40 |
+| 控制提取与缓存 | 80 |
+| D4 degradation screening | 180 |
+| Gate-only | 160 |
+| LoRA、teacher 与消融 | 320 |
+| 最终生成和 evaluator | 180 |
+| 重跑与 contingency | 140 |
+| 总计 | 1,100 |
+
+存储：
+
+- 权重与基础 evaluator：<150GB。
+- 训练视频和控制 cache：0.5–1TB。
+- 多 seed 生成视频：1–2TB。
+- 总 scratch：2–4TB。
+
+不保存全部 latent，只保存 mp4、输入控制、配置、逐样本 JSON 和代表性失败案例。
+
+---
+
+## 12. 风险与切换策略
+
+### 风险 A：污染没有显著影响
+
+判断：D4 中下降 <15% 或小于 seed 方差。  
+处理：立即停止，切换 GenLeak，不扩大数据或训练。
+
+### 风险 B：普通 augmentation 已经足够
+
+判断：CtrlGuard 与 augmentation-only LoRA 差异不显著。  
+处理：不能声称 gate 有算法贡献；可转为 benchmark/analysis 稿，或停止。
+
+### 风险 C：解析 confidence 不可靠
+
+处理顺序：
+
+1. 先使用已知 synthetic corruption mask 验证 gate 上限。
+2. 再训练轻量 reliability predictor。
+3. 不在两周内引入大型 VLM 判断控制质量。
+
+### 风险 D：Gate 全部趋近 0
+
+处理：
+
+- 增强 clean-preservation loss。
+- 保证至少 30% batch 使用 clean control。
+- 报告 control adherence，防止模型通过忽略控制作弊。
+
+### 风险 E：1.3B 生成随机性太大
+
+处理：
+
+- paired seed。
+- 每个结果至少 3 seeds。
+- 优先看相对 control error，不只看审美分数。
+
+### 风险 F：GPU 被其他任务占用
+
+当前 8 张卡被 `/public/hubin/wangmingxin/.../spiderwam/bin/python` 的 16-GPU/2-node 训练占满；不是本项目进程，不能强制结束。  
+本项目使用等待脚本，卡空闲后先跑单卡 smoke，成功后自动拉起 8 卡 baseline。
+
+---
+
+## 13. 投稿定位
+
+### 两周后只有以下结果
+
+- 单个 1.3B 基座。
+- synthetic corruption。
+- 自动指标。
+
+定位：workshop、short paper 或先发技术报告，不应直接包装成主会完整稿。
+
+### 达到主会级证据
+
+- 三种控制模态。
+- synthetic + real extractor shift。
+- 未见 corruption family 泛化。
+- 第二模型尺度或第二基座。
+- 完整消融与人评。
+- clean quality 无显著下降。
+
+可考虑 ICLR/CVPR 主会；最终 venue 应按当时 CFP 和实验成熟度确定。
+
+---
+
+## 14. 当前工程状态
+
+### 已完成
+
+- `Wan-AI/Wan2.1-VACE-1.3B` 权重已下载。
+- `VACE-Benchmark` 已下载。
+- 已克隆 VACE、Wan2.1、DiffSynth-Studio、VBench、VideoScore2、VideoAlign、CRONOS。
+- 已建立隔离 Python 环境 `.venv`，复用系统 CUDA 12.8 / PyTorch 2.7 / FlashAttention。
+- 已编写 8 卡独立数据并行 launcher：
+  - `scripts/launch_baseline_8gpu.py`
+  - `scripts/wait_and_bootstrap_8gpu.sh`
+
+### 八卡启动协议
+
+1. 每张卡空闲显存至少 60GB。
+2. GPU0 先运行 17-frame/8-step 单卡端到端 smoke。
+3. 成功后 8 张卡分别运行 depth、flow、pose、scribble、layout、gray、first-frame、inpainting。
+4. 每个 worker 独立日志、seed、输出目录和 manifest。
+5. 任一 worker 失败则整体标记失败，不悄悄跳过。
+
+### 关键路径
+
+- 论文计划：`PAPER_PLAN.md`
+- 原调研：`RESEARCH_PLAN.md`
+- 模型：`models/Wan2.1-VACE-1.3B`
+- Benchmark：`data/VACE-Benchmark`
+- 输出：`outputs/baseline`
+- Runtime requirements：`requirements-runtime.txt`
+- 一键环境搭建：`scripts/setup_runtime.sh`
diff --git a/video_gen_14d/RESEARCH_PLAN.md b/video_gen_14d/RESEARCH_PLAN.md
new file mode 100644
index 0000000000000000000000000000000000000000..2a559962fedcdc4c0f8e7425c854f7a03eece6d1
--- /dev/null
+++ b/video_gen_14d/RESEARCH_PLAN.md
@@ -0,0 +1,225 @@
+# 14 天视频生成研究计划（2026-07-14）
+
+## 结论
+
+推荐主线：**CtrlGuard：面向污染与冲突条件的可靠性校准视频控制**。
+
+核心问题不是再训练一个更大的视频模型，而是研究 VACE 在真实使用中遇到的结构控制信号不可靠问题：深度、光流、姿态、mask、scribble 可能有局部错误、缺帧、延迟、尺度漂移或彼此冲突。现有 VACE 默认以固定强度注入控制 hint；CtrlGuard 将其改为按层、按时间、按空间 token 的置信门控，并用干净控制教师一致性进行小参数后训练。
+
+这是当前约束下期望收益最高的选择：
+
+- 使用已发布的 Wan2.1-VACE-1.3B，不做基础模型预训练。
+- VACE 权重、官方推理代码、DiffSynth 训练代码均真实可用且为 Apache-2.0。
+- 8×H200 140GB 足够并行生成、评测以及 rank-16/32 LoRA 或小型 gate 训练。
+- 方法有直观可视化，失败也能留下一个可复用的控制污染 benchmark。
+- 14 天可形成强原型、完整消融和论文初稿；完整主会投稿仍建议后续补第二基座和更大规模人评。
+
+## 前沿趋势与开源核验
+
+核验日期：2026-07-14。这里的“开源”要求仓库中存在实际代码且有明确许可证；只有项目页、README 或 “coming soon” 不算。
+
+| 方向 | 工作 | 代码/权重状态 | 许可证与结论 |
+|---|---|---|---|
+| 统一控制/编辑 | [VACE](https://github.com/ali-vilab/VACE) | 官方推理、预处理和 demo 完整；1.3B/14B 权重已发布；官方无训练入口 | 代码与 Wan 权重 Apache-2.0；可直接依赖 |
+| VACE 训练 | [DiffSynth-Studio](https://github.com/modelscope/DiffSynth-Studio) | 有 VACE-1.3B/14B 的 LoRA、全参训练与验证脚本 | Apache-2.0；可直接依赖 |
+| 相机控制 | [UCPE](https://github.com/chengzhag/UCPE) | 完整训练、推理、评测与 PanShot 数据；权重通过 OneDrive | 代码 MIT、数据 CC-BY-4.0；可用，但相机/物体解耦赛道已拥挤 |
+| 轨迹控制 | [MotionPro](https://github.com/HiDream-ai/MotionPro) | 完整训练、数据构造、评测和 HF 权重 | 项目 Apache-2.0；SVD 基座另受 Stability 条款约束 |
+| 因果少步生成 | [Causal Forcing](https://github.com/thu-ml/Causal-Forcing) | 完整训练/推理，1/2/4-step 和长视频权重已发布 | Apache-2.0；活跃，但完整复现成本高 |
+| 流式长视频 | [LongLive](https://github.com/NVlabs/LongLive) | 训练/推理、FP8/NVFP4、量化 KV、流式 VAE 和 5B 权重均已发布 | 代码 Apache-2.0，权重为 NVIDIA Open Model License |
+| 长视频记忆 | [MemRoPE](https://github.com/YoungRaeKimm/MemRoPE) | 实际推理代码已发布；复用 Self-Forcing/LongLive 权重 | Apache-2.0；training-free，可复现 |
+| KV 剪枝 | [Forcing-KV](https://github.com/zju-jiyicheng/Forcing-KV) | 实际推理代码已发布；复用已有因果模型权重 | Apache-2.0；H200 很适合，但方向已快速拥挤 |
+| 视频奖励 | [VideoAlign](https://github.com/KlingAIResearch/VideoAlign) | VideoReward 训练、推理、评测代码和 2B 权重可用；论文中的视频 DPO 并未完整发布 | 代码 MIT、权重 Apache-2.0；可直接作为快速 evaluator |
+| 综合评测 | [VideoScore2](https://github.com/TIGER-AI-Lab/VideoScore2) | 推理、SFT、GRPO、评测代码与 8B 权重已发布 | 代码 MIT、权重 Apache-2.0；适合离线最终评测 |
+| 物理评测 | [CRONOS](https://github.com/GenIntel/CRONOS-benchmark) | 六项评测代码和 18.8GB/675 序列数据已发布 | 代码 MIT、数据 Apache-2.0；可作为高风险备选课题 |
+| 人体物理后训练 | [PhyMotion](https://github.com/h6kplus/PhyMotion) | MuJoCo reward、RL 训练、推理、LoRA 与 prompts 均发布 | 代码 MIT、LoRA Apache-2.0；上游 SMPL/MotionX 条款需另核对 |
+
+不应作为本项目依赖：
+
+- [Quant VideoGen](https://github.com/svg-project/Quant-VideoGen)：有实际代码，但仓库没有许可证；只能称 source-available，不能直接复制进投稿代码。
+- [Flash-GRPO](https://github.com/Shredded-Pork/Flash-GRPO)：有 8 卡训练脚本，但仓库没有许可证。
+- [Self-Forcing++](https://github.com/justincui03/Self-Forcing-Plus-Plus)：截至核验日只有说明和展示资源，代码/权重仍未发布。
+- [LatSearch](https://github.com/zengqunzhao/LatSearch)：截至核验日仍是 README/项目页，Code 与 checkpoint 未发布。
+
+### 目前真正前沿的四条线
+
+1. **双向扩散转向因果/自回归视频生成**：Self/Causal Forcing、LongLive 等在做流式生成和训练—推理分布对齐。
+2. **效率从“少步”扩展到整栈优化**：稀疏注意力、KV 剪枝/量化、FP8、异步 VAE 和蒸馏开始联合设计。
+3. **控制从单条件转向组合控制与可靠性**：相机、轨迹、身份、首尾帧已经拥挤，但“条件本身不可靠时怎么办”仍缺少系统 benchmark 和通用修复。
+4. **后训练从单一美学奖励转向多维质量和物理一致性**：VideoAlign、VideoScore2、PhyMotion、WMReward 等推动 reward/RL/test-time scaling，同时暴露 reward hacking 和泛化问题。
+
+## 候选课题排序
+
+评分考虑新颖性、14 天成功率、开源完整度和可形成证据的强度。
+
+| 排名 | 课题 | 新颖性 | 14 天可行性 | 预计 GPU-hours | 主要风险 |
+|---|---|---:|---:|---:|---|
+| 1 | CtrlGuard：污染/冲突控制的置信门控 | 4/5 | 5/5 | 700–1100 | 退化不显著，或简单 corruption augmentation 已足够 |
+| 2 | GenLeak：去除 reward model 的生成器指纹 | 4/5 | 4/5 | 650–1000 | 需要额外人评，且可能观察不到足够大的 generator bias |
+| 3 | CRONOS 反事实物理一致性 LoRA | 5/5 | 3/5 | 1100–1600 | 只有 675 个合成序列，容易被质疑对 benchmark 过拟合 |
+| 4 | 压缩 memory-in-the-loop 因果蒸馏 | 4/5 | 3/5 | 800–1400 | 最新 KV memory 工作密集，工程与 novelty 风险都高 |
+| 5 | 悲观式、抗 reward hacking 的视频 test-time scaling | 3/5 | 5/5 | 600–1000 | 必须先观察到大 N 下真实 Goodhart 反转，否则题目不成立 |
+
+相机—物体轨迹解耦、通用视频 GRPO、普通 LoRA 融合、单纯 KV 量化和少步蒸馏不建议作为主线：这些方向到 2026 年中已经有密集近邻工作，两周很难建立足够差异。
+
+## CtrlGuard 技术方案
+
+### 研究假设
+
+VACE 对干净结构控制很强，但固定 `vace_scale` 无法区分可靠与不可靠 token。真实提取器的误差会被 15 层 VACE hint 持续注入主干，导致控制偏差、结构变形和时序闪烁。
+
+当前 DiffSynth 实现中，每个 VACE layer 的 hint 以标量强度注入：
+
+`x = x + current_vace_hint * vace_scale`
+
+CtrlGuard 将其替换为：
+
+`x = x + current_vace_hint * g(layer, timestep, token, condition)`
+
+其中 `g ∈ [0, 1]` 是小型可靠性网络输出的时空门控。
+
+### 最小方法
+
+1. 冻结 Wan2.1-VACE-1.3B。
+2. 将控制视频、解析置信度图、hint 局部统计、denoising timestep 和 layer id 输入小型 3D conv + MLP gate。
+3. 对干净条件使用 `g≈1` 约束，保证 clean performance。
+4. 对随机污染条件使用：
+   - flow-matching/SFT loss；
+   - 干净控制 teacher 的 noise-prediction 或 hint consistency；
+   - 已知合成污染 mask/severity 的 gate calibration loss。
+5. 第一阶段只训练约 0.5–3M gate；若恢复能力不足，再增加 VACE `after_proj` 或 q/k/v 的 rank-16 LoRA。
+
+### 污染族
+
+- 空间：随机孔洞、局部遮挡、边缘侵蚀/膨胀、模糊、量化和噪声。
+- 时间：丢帧、重复帧、局部延迟、时间抖动、速度漂移。
+- 几何：深度 scale/shift、光流 bias、姿态关键点偏移或缺失、mask 边界漂移。
+- 真实 OOD：不同 depth/flow/pose 提取器之间的误差与置信度。
+- 冲突：两种控制在局部区域或时间段给出不一致指示；作为 stretch goal，不阻塞最小论文。
+
+### 数据
+
+- **评测只用**：VACE-Benchmark（Apache-2.0，2.1GB），按 source video 隔离，避免同源泄漏。
+- **训练 smoke test**：VACE-Benchmark 的非评测子集，只用于 D1–D5 判定方法是否值得继续。
+- **正式训练**：抽取约 2,000–5,000 个 OpenVid-1M 视频片段，离线生成 depth/flow/pose 控制。OpenVid 卡片标 CC-BY-4.0 且限定研究用途，并要求遵守上游视频许可证；投稿时只发布索引、派生控制和脚本，不重新分发原视频。
+- 所有 clean/corrupt pair 使用同一视频、prompt、seed 和 timestep，减少生成随机性。
+
+### 指标
+
+控制遵循：
+
+- Depth：scale-aligned AbsRel / RMSE。
+- Flow：EPE 与 temporal warping error。
+- Pose：PCK / OKS、关键点丢失率。
+- Mask/layout：IoU、Boundary-F。
+- 汇总：相对性能下降、mCE（mean Corruption Error）、severity AUC、worst-group score。
+
+生成质量：
+
+- VBench 的 subject/background consistency、motion smoothness、dynamic degree。
+- VideoScore2 的 visual quality、text alignment、physical/common-sense consistency。
+- VideoAlign VQ/MQ/TA 作为快速筛选，不作为唯一最终指标。
+- 至少 100–200 对、3 人盲评；评测样本和训练 reward/teacher 分离。
+
+可靠性：
+
+- gate 与真实控制误差的 AUROC/AUPRC。
+- Expected Calibration Error。
+- clean 条件下质量退化必须不超过 2%。
+
+### 必做基线
+
+1. 原始 VACE，固定 `vace_scale=1`。
+2. 最优全局 `vace_scale`。
+3. Control dropout。
+4. Corruption augmentation + 普通 LoRA。
+5. 解析 confidence map 的 training-free mask。
+6. CtrlGuard gate-only。
+7. CtrlGuard gate + rank-16 LoRA。
+
+## 14 天执行表
+
+| 日期 | 工作 | 当日交付与门槛 |
+|---|---|---|
+| D1 | 建环境，跑通 VACE-1.3B clean inference；实测 H200 吞吐/显存 | 24 个 clean 样本成功；固定 seed 可复现 |
+| D2 | 整理 VACE-Benchmark split；实现 depth/flow/pose 基础指标 | 自动指标与人工 spot-check 方向一致 |
+| D3 | 实现 10–12 种合成污染和 5 档 severity | 每种污染都有单元样例与参数记录 |
+| D4 | 运行原始 VACE degradation curves | 中等污染使至少两类控制下降 ≥15%，且效应大于跨 seed 方差 |
+| D5 | 加入跨提取器 OOD；完成第一个 Go/No-Go | 若 D4 门槛不成立，立即切换 GenLeak，不继续训练 |
+| D6 | 在 DiffSynth 中实现 per-token/per-layer gate | gate=1 时数值上复现原模型 |
+| D7 | gate-only 训练与小规模验证 | 恢复至少 20% 的损失，clean 降幅 ≤2% |
+| D8 | rank-16 LoRA 与 teacher consistency | 相比 augmentation-only 有明确增益 |
+| D9 | gate 输入、layer/time conditioning、loss 消融 | 锁定最终配置，不再扩方法 |
+| D10 | 全量 corruption benchmark，3 seeds | 主表完成，保存逐样本结果 |
+| D11 | VBench、VideoScore2、VideoAlign；失败案例分类 | held-out evaluator 不出现反向退化 |
+| D12 | 100–200 对盲评；可选 14B corruption transfer | 人评协议、置信区间和一致率完成 |
+| D13 | 统计显著性、图表、supplement 与代码清理 | 所有结论能从脚本一键复算 |
+| D14 | 论文初稿与最终 Go/No-Go | 达标则准备主会扩展；否则整理 workshop/benchmark 稿 |
+
+## 资源预算
+
+硬件理论上限：`8 GPUs × 24 h × 14 d = 2,688 H200 GPU-hours`。
+
+建议主项目上限为 1,100 GPU-hours，保留约 1,588 GPU-hours 给失败重跑、第二模型或后续扩展：
+
+| 项目 | GPU-hours 上限 |
+|---|---:|
+| 环境、smoke test、吞吐标定 | 40 |
+| 控制提取与 corruption cache | 80 |
+| 原始模型 degradation benchmark | 180 |
+| gate-only 训练与搜索 | 160 |
+| LoRA、teacher consistency 和消融 | 320 |
+| VBench/VideoScore2/VideoAlign/人评样本生成 | 180 |
+| 最终重跑与 contingency | 140 |
+| 合计 | 1,100 |
+
+存储建议：2–4TB scratch。模型和 benchmark 本身很小，主要空间来自多 seed、多 severity 的生成视频和评测缓存。不要长期保存所有 latent；保留 mp4、控制图、配置、逐样本 JSON 和必要的失败案例。
+
+## Go / No-Go
+
+### D4 必须满足
+
+- 中等污染造成至少 15% 的控制指标下降；
+- 污染效应显著大于 seed 方差；
+- 至少 depth/flow/pose 中两类成立。
+
+不满足则说明问题不够强，立即切换到 GenLeak 奖励模型泛化课题。
+
+### D8 必须满足
+
+- CtrlGuard 恢复至少 30% 的控制损失；
+- clean 指标下降不超过 2%；
+- 明显优于只做 corruption augmentation 的 LoRA。
+
+不满足则不应包装成算法论文；最多保留为 benchmark/negative result。
+
+### 主会级最低证据
+
+- 三种控制模态、六种以上污染、合成与真实提取器误差；
+- 未见 corruption family 和未见 extractor 的 OOD 泛化；
+- gate-only、LoRA、augmentation、training-free confidence 的完整消融；
+- 至少一个第二模型尺度或基座上的验证；
+- 自动指标与盲评方向一致。
+
+只有单基座、单 benchmark 时，更现实的定位是 workshop 或短文；达到上述证据后再考虑 ICLR/CVPR 主会。
+
+## 当前本地资产
+
+工作目录：`/public/interns/hubin/jmtseng/video_gen_14d`
+
+已克隆并固定当前 revision：
+
+| 目录 | revision | 日期 | 用途 |
+|---|---|---|---|
+| `third_party/VACE` | `48eb44f` | 2025-10-17 | 官方推理与控制预处理 |
+| `third_party/DiffSynth-Studio` | `fb337fb` | 2026-07-14 | VACE LoRA/全参训练 |
+| `third_party/VBench` | `45e79ec` | 2026-03-23 | 生成质量评测 |
+| `third_party/VideoScore2` | `a88168a` | 2025-12-08 | 综合离线 evaluator |
+| `third_party/VideoAlign` | `219ab9d` | 2025-09-24 | 快速视频 reward |
+| `third_party/CRONOS-benchmark` | `c56e317` | 2026-05-27 | 高风险物理一致性备选 |
+
+主模型 `Wan-AI/Wan2.1-VACE-1.3B` 共 19.0GB，VACE-Benchmark 共 2.1GB，下载目录分别为：
+
+- `models/Wan2.1-VACE-1.3B`
+- `data/VACE-Benchmark`
+
+Hugging Face 下载工具安装在隔离环境 `.venv-hf`，不影响后续训练环境。
diff --git a/video_gen_14d/data/VACE-Benchmark/.gitattributes b/video_gen_14d/data/VACE-Benchmark/.gitattributes
new file mode 100644
index 0000000000000000000000000000000000000000..1ef325f1b111266a6b26e0196871bd78baa8c2f3
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/.gitattributes
@@ -0,0 +1,59 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.lz4 filter=lfs diff=lfs merge=lfs -text
+*.mds filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+# Audio files - uncompressed
+*.pcm filter=lfs diff=lfs merge=lfs -text
+*.sam filter=lfs diff=lfs merge=lfs -text
+*.raw filter=lfs diff=lfs merge=lfs -text
+# Audio files - compressed
+*.aac filter=lfs diff=lfs merge=lfs -text
+*.flac filter=lfs diff=lfs merge=lfs -text
+*.mp3 filter=lfs diff=lfs merge=lfs -text
+*.ogg filter=lfs diff=lfs merge=lfs -text
+*.wav filter=lfs diff=lfs merge=lfs -text
+# Image files - uncompressed
+*.bmp filter=lfs diff=lfs merge=lfs -text
+*.gif filter=lfs diff=lfs merge=lfs -text
+*.png filter=lfs diff=lfs merge=lfs -text
+*.tiff filter=lfs diff=lfs merge=lfs -text
+# Image files - compressed
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text
+*.webp filter=lfs diff=lfs merge=lfs -text
+# Video files - compressed
+*.mp4 filter=lfs diff=lfs merge=lfs -text
+*.webm filter=lfs diff=lfs merge=lfs -text
diff --git a/video_gen_14d/data/VACE-Benchmark/README.md b/video_gen_14d/data/VACE-Benchmark/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..180c8f048a3a4ea06348a26da4e01ae5b1712085
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/README.md
@@ -0,0 +1,203 @@
+---
+license: apache-2.0
+language:
+- en
+- zh
+tags:
+- art
+size_categories:
+- n<1K
+---
+<p align="center">
+
+<h1 align="center">VACE: All-in-One Video Creation and Editing</h1>
+<h3 align="center">(ICCV 2025)</h3>
+<p align="center">
+    <strong>Zeyinzi Jiang<sup>*</sup></strong>
+    ·
+    <strong>Zhen Han<sup>*</sup></strong>
+    ·
+    <strong>Chaojie Mao<sup>*&dagger;</sup></strong>
+    ·
+    <strong>Jingfeng Zhang</strong>
+    ·
+    <strong>Yulin Pan</strong>
+    ·
+    <strong>Yu Liu</strong>
+    <br>
+    <b>Tongyi Lab - <a href="https://github.com/Wan-Video/Wan2.1"><img src='https://ali-vilab.github.io/VACE-Page/assets/logos/wan_logo.png' alt='wan_logo' style='margin-bottom: -4px; height: 20px;'></a> </b>
+    <br>
+    <br>
+        <a href="https://arxiv.org/abs/2503.07598"><img src='https://img.shields.io/badge/VACE-arXiv-red' alt='Paper PDF'></a>
+        <a href="https://ali-vilab.github.io/VACE-Page/"><img src='https://img.shields.io/badge/VACE-Project_Page-green' alt='Project Page'></a>
+        <a href="https://huggingface.co/collections/ali-vilab/vace-67eca186ff3e3564726aff38"><img src='https://img.shields.io/badge/VACE-HuggingFace_Model-yellow'></a>
+        <a href="https://modelscope.cn/collections/VACE-8fa5fcfd386e43"><img src='https://img.shields.io/badge/VACE-ModelScope_Model-purple'></a>
+    <br>
+</p>
+
+
+## Introduction
+<strong>VACE</strong> is an all-in-one model designed for video creation and editing. It encompasses various tasks, including reference-to-video generation (<strong>R2V</strong>), video-to-video editing (<strong>V2V</strong>), and masked video-to-video editing (<strong>MV2V</strong>), allowing users to compose these tasks freely. This functionality enables users to explore diverse possibilities and streamlines their workflows effectively, offering a range of capabilities, such as Move-Anything, Swap-Anything, Reference-Anything, Expand-Anything, Animate-Anything, and more.
+
+<img src='https://raw.githubusercontent.com/ali-vilab/VACE/refs/heads/main/assets/materials/teaser.jpg'>
+
+
+## 🎉 News
+- [x] Oct 17, 2025: [VACE-Benchmark](https://huggingface.co/datasets/ali-vilab/VACE-Benchmark) has been updated to incorporate the evaluation data. [VACE-Page](https://ali-vilab.github.io/VACE-Page/) also features creative community cases, offering researchers and community members better project insight and tracking.
+- [x] Jun 26, 2025: [VACE](https://openaccess.thecvf.com/content/ICCV2025/html/Jiang_VACE_All-in-One_Video_Creation_and_Editing_ICCV_2025_paper.html) is accepted by ICCV 2025.
+- [x] May 14, 2025: 🔥Wan2.1-VACE-1.3B and Wan2.1-VACE-14B models are now available at [HuggingFace](https://huggingface.co/collections/Wan-AI/wan21-68ac4ba85372ae5a8e282a1b) and [ModelScope](https://modelscope.cn/collections/tongyiwanxiang-Wan21-shipinshengcheng-67ec9b23fd8d4f)!
+- [x] Mar 31, 2025: 🔥VACE-Wan2.1-1.3B-Preview and VACE-LTX-Video-0.9 models are now available at [HuggingFace](https://huggingface.co/collections/ali-vilab/vace-67eca186ff3e3564726aff38) and [ModelScope](https://modelscope.cn/collections/VACE-8fa5fcfd386e43)!
+- [x] Mar 31, 2025: 🔥Release code of model inference, preprocessing, and gradio demos. 
+- [x] Mar 11, 2025: We propose [VACE](https://ali-vilab.github.io/VACE-Page/), an all-in-one model for video creation and editing.
+
+
+## 🪄 Models
+| Models                   | Download Link                                                                                                                                           | Video Size        | License                                                                                       |
+|--------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------|-----------------------------------------------------------------------------------------------|
+| VACE-Wan2.1-1.3B-Preview | [Huggingface](https://huggingface.co/ali-vilab/VACE-Wan2.1-1.3B-Preview) 🤗  [ModelScope](https://modelscope.cn/models/iic/VACE-Wan2.1-1.3B-Preview) 🤖 | ~ 81 x 480 x 832  | [Apache-2.0](https://huggingface.co/Wan-AI/Wan2.1-T2V-1.3B/blob/main/LICENSE.txt)             |
+| VACE-LTX-Video-0.9       | [Huggingface](https://huggingface.co/ali-vilab/VACE-LTX-Video-0.9) 🤗     [ModelScope](https://modelscope.cn/models/iic/VACE-LTX-Video-0.9) 🤖          | ~ 97 x 512 x 768  | [RAIL-M](https://huggingface.co/Lightricks/LTX-Video/blob/main/ltx-video-2b-v0.9.license.txt) |
+| Wan2.1-VACE-1.3B         | [Huggingface](https://huggingface.co/Wan-AI/Wan2.1-VACE-1.3B) 🤗     [ModelScope](https://www.modelscope.cn/models/Wan-AI/Wan2.1-VACE-1.3B) 🤖          | ~ 81 x 480 x 832  | [Apache-2.0](https://huggingface.co/Wan-AI/Wan2.1-T2V-1.3B/blob/main/LICENSE.txt)             |
+| Wan2.1-VACE-14B          | [Huggingface](https://huggingface.co/Wan-AI/Wan2.1-VACE-14B) 🤗     [ModelScope](https://www.modelscope.cn/models/Wan-AI/Wan2.1-VACE-14B) 🤖            | ~ 81 x 720 x 1280 | [Apache-2.0](https://huggingface.co/Wan-AI/Wan2.1-T2V-14B/blob/main/LICENSE.txt)             |
+
+- The input supports any resolution, but to achieve optimal results, the video size should fall within a specific range.
+- All models inherit the license of the original model.
+
+
+## ⚙️ Installation
+The codebase was tested with Python 3.10.13, CUDA version 12.4, and PyTorch >= 2.5.1.
+
+### Setup for Model Inference
+You can setup for VACE model inference by running:
+```bash
+git clone https://github.com/ali-vilab/VACE.git && cd VACE
+pip install torch==2.5.1 torchvision==0.20.1 --index-url https://download.pytorch.org/whl/cu124  # If PyTorch is not installed.
+pip install -r requirements.txt
+pip install wan@git+https://github.com/Wan-Video/Wan2.1  # If you want to use Wan2.1-based VACE.
+pip install ltx-video@git+https://github.com/Lightricks/LTX-Video@ltx-video-0.9.1 sentencepiece --no-deps # If you want to use LTX-Video-0.9-based VACE. It may conflict with Wan.
+```
+Please download your preferred base model to `<repo-root>/models/`. 
+
+### Setup for Preprocess Tools
+If you need preprocessing tools, please install:
+```bash
+pip install -r requirements/annotator.txt
+```
+Please download [VACE-Annotators](https://huggingface.co/ali-vilab/VACE-Annotators) to `<repo-root>/models/`.
+
+### Local Directories Setup
+It is recommended to download [VACE-Benchmark](https://huggingface.co/datasets/ali-vilab/VACE-Benchmark) to `<repo-root>/benchmarks/` as examples in `run_vace_xxx.sh`.
+
+We recommend to organize local directories as:
+```angular2html
+VACE
+├── ...
+├── benchmarks
+│   └── VACE-Benchmark
+│       └── assets
+│           └── examples
+│               ├── animate_anything
+│               │   └── ...
+│               └── ...
+├── models
+│   ├── VACE-Annotators
+│   │   └── ...
+│   ├── VACE-LTX-Video-0.9
+│   │   └── ...
+│   └── VACE-Wan2.1-1.3B-Preview
+│       └── ...
+└── ...
+```
+
+## 🚀 Usage
+In VACE, users can input **text prompt** and optional **video**, **mask**, and **image** for video generation or editing.
+Detailed instructions for using VACE can be found in the [User Guide](./UserGuide.md).
+
+### Inference CIL
+#### 1) End-to-End Running
+To simply run VACE without diving into any implementation details, we suggest an end-to-end pipeline. For example:
+```bash
+# run V2V depth
+python vace/vace_pipeline.py --base wan --task depth --video assets/videos/test.mp4 --prompt 'xxx'
+
+# run MV2V inpainting by providing bbox
+python vace/vace_pipeline.py --base wan --task inpainting --mode bbox --bbox 50,50,550,700 --video assets/videos/test.mp4 --prompt 'xxx'
+```
+This script will run video preprocessing and model inference sequentially, 
+and you need to specify all the required args of preprocessing (`--task`, `--mode`, `--bbox`, `--video`, etc.) and inference (`--prompt`, etc.). 
+The output video together with intermediate video, mask and images will be saved into `./results/` by default.
+
+> 💡**Note**:
+> Please refer to [run_vace_pipeline.sh](https://github.com/ali-vilab/VACE/blob/main/run_vace_pipeline.sh) for usage examples of different task pipelines.
+
+
+#### 2) Preprocessing
+To have more flexible control over the input, before VACE model inference, user inputs need to be preprocessed into `src_video`, `src_mask`, and `src_ref_images` first.
+We assign each [preprocessor](https://raw.githubusercontent.com/ali-vilab/VACE/refs/heads/main/vace/configs/__init__.py) a task name, so simply call [`vace_preprocess.py`](https://raw.githubusercontent.com/ali-vilab/VACE/refs/heads/main/vace/vace_preproccess.py) and specify the task name and task params. For example:
+```angular2html
+# process video depth
+python vace/vace_preproccess.py --task depth --video assets/videos/test.mp4
+
+# process video inpainting by providing bbox
+python vace/vace_preproccess.py --task inpainting --mode bbox --bbox 50,50,550,700 --video assets/videos/test.mp4
+```
+The outputs will be saved to `./processed/` by default.
+
+> 💡**Note**:
+> Please refer to [run_vace_pipeline.sh](https://github.com/ali-vilab/VACE/blob/main/run_vace_pipeline.sh) preprocessing methods for different tasks.
+Moreover, refer to [vace/configs/](https://github.com/ali-vilab/VACE/blob/main/vace/configs/) for all the pre-defined tasks and required params.
+You can also customize preprocessors by implementing at [`annotators`](https://github.com/ali-vilab/VACE/blob/main/vace/annotators/__init__.py) and register them at [`configs`](https://github.com/ali-vilab/VACE/blob/main/vace/configs).
+
+
+#### 3) Model inference
+Using the input data obtained from **Preprocessing**, the model inference process can be performed as follows:
+```bash
+# For Wan2.1 single GPU inference (1.3B-480P)
+python vace/vace_wan_inference.py --ckpt_dir <path-to-model> --src_video <path-to-src-video> --src_mask <path-to-src-mask> --src_ref_images <paths-to-src-ref-images> --prompt "xxx"
+
+# For Wan2.1 Multi GPU Acceleration inference (1.3B-480P)
+pip install "xfuser>=0.4.1"
+torchrun --nproc_per_node=8 vace/vace_wan_inference.py --dit_fsdp --t5_fsdp --ulysses_size 1 --ring_size 8 --ckpt_dir <path-to-model> --src_video <path-to-src-video> --src_mask <path-to-src-mask> --src_ref_images <paths-to-src-ref-images> --prompt "xxx"
+
+# For Wan2.1 Multi GPU Acceleration inference (14B-720P)
+torchrun --nproc_per_node=8 vace/vace_wan_inference.py --dit_fsdp --t5_fsdp --ulysses_size 8 --ring_size 1 --size 720p --model_name 'vace-14B' --ckpt_dir <path-to-model> --src_video <path-to-src-video> --src_mask <path-to-src-mask> --src_ref_images <paths-to-src-ref-images> --prompt "xxx"
+
+# For LTX inference, run
+python vace/vace_ltx_inference.py --ckpt_path <path-to-model> --text_encoder_path <path-to-model> --src_video <path-to-src-video> --src_mask <path-to-src-mask> --src_ref_images <paths-to-src-ref-images> --prompt "xxx"
+```
+The output video together with intermediate video, mask and images will be saved into `./results/` by default.
+
+> 💡**Note**: 
+> (1) Please refer to [vace/vace_wan_inference.py](https://github.com/ali-vilab/VACE/blob/main/vace/vace_wan_inference.py) and [vace/vace_ltx_inference.py](https://github.com/ali-vilab/VACE/blob/main/vace/vace_ltx_inference.py) for the inference args.
+> (2) For LTX-Video and English language Wan2.1 users, you need prompt extension to unlock the full model performance. 
+Please follow the [instruction of Wan2.1](https://github.com/Wan-Video/Wan2.1?tab=readme-ov-file#2-using-prompt-extension) and set `--use_prompt_extend` while running inference.
+> (3) When performing prompt extension in editing tasks, it's important to pay attention to the results of expanding plain text. Since the visual information being input is unknown, this may lead to the extended output not matching the video being edited, which can affect the final outcome.
+
+### Inference Gradio
+For preprocessors, run 
+```bash
+python vace/gradios/vace_preprocess_demo.py
+```
+For model inference, run
+```bash
+# For Wan2.1 gradio inference
+python vace/gradios/vace_wan_demo.py
+
+# For LTX gradio inference
+python vace/gradios/vace_ltx_demo.py
+```
+
+## Acknowledgement
+
+We are grateful for the following awesome projects, including [Scepter](https://github.com/modelscope/scepter), [Wan](https://github.com/Wan-Video/Wan2.1), and [LTX-Video](https://github.com/Lightricks/LTX-Video). Additionally, we extend our deepest gratitude to all community creators. It is their proactive exploration, experimentation, and boundless creativity that have brought immense inspiration to the project, fostering the emergence of even more refined workflows and stunning video generation content based on it. This includes, but is not limited to: [Kijai's Workflow](https://github.com/kijai/ComfyUI-WanVideoWrapper), native code support for [ComfyUI](https://github.com/comfyanonymous/ComfyUI) and [Diffusers](https://github.com/huggingface/diffusers), crucial model quantization support, a diverse ecosystem of LoRA adapters, and the ever-evolving innovative workflows from our community members.
+
+
+## BibTeX
+
+```bibtex
+@inproceedings{vace,
+    title = {VACE: All-in-One Video Creation and Editing},
+    author = {Jiang, Zeyinzi and Han, Zhen and Mao, Chaojie and Zhang, Jingfeng and Pan, Yulin and Liu, Yu},
+    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision},
+    pages = {17191-17202},
+    year = {2025}
+}
\ No newline at end of file
diff --git a/video_gen_14d/data/VACE-Benchmark/UserGuide.md b/video_gen_14d/data/VACE-Benchmark/UserGuide.md
new file mode 100644
index 0000000000000000000000000000000000000000..9fd7513a127ad13ed5f40046455b43ca2c28bd9d
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/UserGuide.md
@@ -0,0 +1,160 @@
+# VACE User Guide
+
+## 1. Overall Steps
+
+- Preparation: Be aware of the task type ([single task](#32-single-task) or [multi-task composition](#33-composition-task)) of your creative idea, and prepare all the required materials (images, videos, prompt, etc.)
+- Preprocessing: Select the appropriate preprocessing method based task name, then preprocess your materials to meet the model's input requirements.
+- Inference: Based on the preprocessed materials, perform VACE inference to obtain results.
+
+## 2. Preparations
+
+### 2.1 Task Definition
+
+VACE, as a unified video generation solution, simultaneously supports Video Generation, Video Editing, and complex composition task. Specifically:
+
+- Video Generation: No video input. Injecting concepts into the model through semantic understanding based on text and reference materials, including **T2V** (Text-to-Video Generation) and **R2V** (Reference-to-Video Generation) tasks.
+- Video Editing: With video input. Modifying input video at the pixel level globally or locally,including **V2V** (Video-to-Video Editing) and **MV2V** (Masked Video-to-Video Editing).
+- Composition Task: Compose two or more single task above into a complex composition task, such as **Reference Anything** (Face R2V + Object R2V), **Move Anything**(Frame R2V + Layout V2V), **Animate Anything**(R2V + Pose V2V), **Swap Anything**(R2V + Inpainting MV2V), and **Expand Anything**(Object R2V + Frame R2V + Outpainting MV2V), etc.
+
+Single tasks and compositional tasks are illustrated in the diagram below:
+
+![vace_task](https://raw.githubusercontent.com/ali-vilab/VACE/refs/heads/main/assets/materials/tasks.png)
+
+
+### 2.2 Limitations
+
+- Super high resolution video will be resized to proper spatial size.
+- Super long video will be trimmed or uniformly sampled into around 5 seconds. 
+- For users who are demanding of long video generation, we recommend to generate 5s video clips one by one, while using `firstclip` video extension task to keep the temporal consistency.
+
+## 3. Preprocessing
+### 3.1 VACE-Recognizable Inputs
+
+User-collected materials needs to be preprocessed into VACE-recognizable inputs, including **`src_video`**, **`src_mask`**, **`src_ref_images`**, and **`prompt`**. 
+Specific descriptions are as follows:
+
+- `src_video`: The video to be edited for input into the model, such as condition videos (Depth, Pose, etc.) or in/outpainting input video. **Gray areas**(values equal to 127) represent missing video part. In first-frame R2V task, the first frame are reference frame while subsequent frames are left gray. The missing parts of in/outpainting `src_video` are also set gray.
+- `src_mask`: A 3D mask in the same shape of `src_video`. **White areas** represent the parts to be generated, while **black areas** represent the parts to be retained.
+- `src_ref_images`: Reference images of R2V. Salient object segmentation can be performed to keep the background white. 
+- `prompt`: A text describing the content of the output video. Prompt expansion can be used to achieve better generation effects for LTX-Video and English user of Wan2.1. Use descriptive prompt instead of instructions.
+
+Among them, `prompt` is required while `src_video`, `src_mask`, and `src_ref_images` are optional. For instance, MV2V task requires `src_video`, `src_mask`, and `prompt`; R2V task only requires `src_ref_images` and `prompt`. 
+
+### 3.2 Preprocessing Tools
+Both command line and Gradio demo are supported.
+
+1) Command Line: You can refer to the `run_vace_preproccess.sh` script and invoke it based on the different task types. An example command is as follows:
+```bash
+python vace/vace_preproccess.py --task depth --video assets/videos/test.mp4
+```
+
+2) Gradio Interactive: Launch the graphical interface for data preprocessing and perform preprocessing on the interface. The specific command is as follows:
+```bash
+python vace/gradios/preprocess_demo.py
+```
+
+![gr_pre_demo](https://raw.githubusercontent.com/ali-vilab/VACE/refs/heads/main/assets/materials/gr_pre_demo.jpg)
+
+
+### 3.2 Single Tasks
+
+VACE is an all-in-one model supporting various task types. However, different preprocessing is required for these task types. The specific task types and descriptions are as follows:
+
+| Task       | Subtask              | Annotator                  | Input modal                              | Params                                                                                                                                                                                                               | Note                                                   |
+|------------|----------------------|----------------------------|------------------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------|
+| txt2vid    | txt2vid              | /                          | /                                        | /                                                                                                                                                                                                                    |                                                        |
+| control    | depth                | DepthVideoAnnotator        | video                                    | /                                                                                                                                                                                                                    |                                                        |
+| control    | flow                 | FlowVisAnnotator           | video                                    | /                                                                                                                                                                                                                    |                                                        |
+| control    | gray                 | GrayVideoAnnotator         | video                                    | /                                                                                                                                                                                                                    |                                                        |
+| control    | pose                 | PoseBodyFaceVideoAnnotator | video                                    | /                                                                                                                                                                                                                    |                                                        |
+| control    | scribble             | ScribbleVideoAnnotator     | video                                    | /                                                                                                                                                                                                                    |                                                        |
+| control    | layout_bbox          | LayoutBboxAnnotator        | two bboxes <br>'x1,y1,x2,y2 x1,y1,x2,y2' | /                                                                                                                                                                                                                    | Move linearly from the first box to the second box     |
+| control    | layout_track         | LayoutTrackAnnotator       | video                                    | mode='masktrack/bboxtrack/label/caption'<br>maskaug_mode(optional)='original/original_expand/hull/hull_expand/bbox/bbox_expand'<br>maskaug_ratio(optional)=0~1.0                                                     | Mode represents different methods of subject tracking. |
+| extension  | frameref             | FrameRefExpandAnnotator    | image                                    | mode='firstframe'<br>expand_num=80 (default)                                                                                                                                                                         |                                                        |
+| extension  | frameref             | FrameRefExpandAnnotator    | image                                    | mode='lastframe'<br>expand_num=80 (default)                                                                                                                                                                          |                                                        |
+| extension  | frameref             | FrameRefExpandAnnotator    | two images<br>a.jpg,b.jpg                | mode='firstlastframe'<br>expand_num=80 (default)                                                                                                                                                                     | Images are separated by commas.                        |
+| extension  | clipref              | FrameRefExpandAnnotator    | video                                    | mode='firstclip'<br>expand_num=80 (default)                                                                                                                                                                          |                                                        |
+| extension  | clipref              | FrameRefExpandAnnotator    | video                                    | mode='lastclip'<br>expand_num=80 (default)                                                                                                                                                                           |                                                        |
+| extension  | clipref              | FrameRefExpandAnnotator    | two videos<br>a.mp4,b.mp4                | mode='firstlastclip'<br>expand_num=80 (default)                                                                                                                                                                      | Videos are separated by commas.                        |
+| repainting | inpainting_mask      | InpaintingAnnotator        | video                                    | mode='salient'                                                                                                                                                                                                       | Use salient as a fixed mask.                           |
+| repainting | inpainting_mask      | InpaintingAnnotator        | video + mask                             | mode='mask'                                                                                                                                                                                                          | Use mask as a fixed mask.                              |
+| repainting | inpainting_bbox      | InpaintingAnnotator        | video + bbox<br>'x1, y1, x2, y2'         | mode='bbox'                                                                                                                                                                                                          | Use bbox as a fixed mask.                              |
+| repainting | inpainting_masktrack | InpaintingAnnotator        | video                                    | mode='salientmasktrack'                                                                                                                                                                                              | Use salient mask for dynamic tracking.                 |
+| repainting | inpainting_masktrack | InpaintingAnnotator        | video                                    | mode='salientbboxtrack'                                                                                                                                                                                              | Use salient bbox for dynamic tracking.                 |
+| repainting | inpainting_masktrack | InpaintingAnnotator        | video + mask                             | mode='masktrack'                                                                                                                                                                                                     | Use mask for dynamic tracking.                         |
+| repainting | inpainting_bboxtrack | InpaintingAnnotator        | video + bbox<br>'x1, y1, x2, y2'         | mode='bboxtrack'                                                                                                                                                                                                     | Use bbox for dynamic tracking.                         |
+| repainting | inpainting_label     | InpaintingAnnotator        | video + label                            | mode='label'                                                                                                                                                                                                         | Use label for dynamic tracking.                        |
+| repainting | inpainting_caption   | InpaintingAnnotator        | video + caption                          | mode='caption'                                                                                                                                                                                                       | Use caption for dynamic tracking.                      |
+| repainting | outpainting          | OutpaintingVideoAnnotator  | video                                    | direction=left/right/up/down<br>expand_ratio=0~1.0                                                                                                                                                                   | Combine outpainting directions arbitrarily.            |
+| reference  | image_reference      | SubjectAnnotator           | image                                    | mode='salient/mask/bbox/salientmasktrack/salientbboxtrack/masktrack/bboxtrack/label/caption'<br>maskaug_mode(optional)='original/original_expand/hull/hull_expand/bbox/bbox_expand'<br>maskaug_ratio(optional)=0~1.0 | Use different methods to obtain the subject region.    |
+
+### 3.3 Composition Task
+
+Moreover, VACE supports combining tasks to accomplish more complex objectives. The following examples illustrate how tasks can be combined, but these combinations are not limited to the examples provided:
+
+| Task        | Subtask            | Annotator                  | Input modal        | Params                                                                                                                                                           | Note                                                                                                                           |
+|-------------|--------------------|----------------------------|--------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------|
+| composition | reference_anything | ReferenceAnythingAnnotator | image_list         | mode='salientmasktrack/salientbboxtrack/masktrack/bboxtrack/label/caption'                                                                                       | Input no more than three images.                                                                                               |
+| composition | animate_anything   | AnimateAnythingAnnotator   | image + video      | mode='salientmasktrack/salientbboxtrack/masktrack/bboxtrack/label/caption'                                                                                       | Video for conditional redrawing; images for reference generation.                                                              |
+| composition | swap_anything      | SwapAnythingAnnotator      | image + video      | mode='masktrack/bboxtrack/label/caption'<br>maskaug_mode(optional)='original/original_expand/hull/hull_expand/bbox/bbox_expand'<br>maskaug_ratio(optional)=0~1.0 | Video for conditional redrawing; images for reference generation.<br>Comma-separated mode: first for video, second for images. |
+| composition | expand_anything    | ExpandAnythingAnnotator    | image + image_list | mode='masktrack/bboxtrack/label/caption'<br>direction=left/right/up/down<br>expand_ratio=0~1.0<br>expand_num=80 (default)                                        | First image for extension edit; others for reference.<br>Comma-separated mode: first for video, second for images.             |
+| composition | move_anything      | MoveAnythingAnnotator      | image + two bboxes | expand_num=80 (default)                                                                                                                                          | First image for initial frame reference; others represented by linear bbox changes.                                            |
+| composition | more_anything      | ...                        | ...                | ...                                                                                                                                                              | ...                                                                                                                            |
+
+
+## 4. Model Inference
+
+### 4.1 Execution Methods
+
+Both command line and Gradio demo are supported.
+
+1) Command Line: Refer to the `run_vace_ltx.sh` and `run_vace_wan.sh` scripts and invoke them based on the different task types. The input data needs to be preprocessed to obtain parameters such as `src_video`, `src_mask`, `src_ref_images` and `prompt`. An example command is as follows:
+```bash
+python vace/vace_wan_inference.py --src_video <path-to-src-video> --src_mask <path-to-src-mask> --src_ref_images <paths-to-src-ref-images> --prompt <prompt>  # wan
+python vace/vace_ltx_inference.py --src_video <path-to-src-video> --src_mask <path-to-src-mask> --src_ref_images <paths-to-src-ref-images> --prompt <prompt>  # ltx
+```
+
+2) Gradio Interactive: Launch the graphical interface for model inference and perform inference through interactions on the interface. The specific command is as follows:
+```bash
+python vace/gradios/vace_wan_demo.py  # wan
+python vace/gradios/vace_ltx_demo.py  # ltx
+```
+
+![gr_infer_demo](https://raw.githubusercontent.com/ali-vilab/VACE/refs/heads/main/assets/materials/gr_infer_demo.jpg)
+
+3) End-to-End Inference: Refer to the `run_vace_pipeline.sh` script and invoke it based on different task types and input data. This pipeline includes both preprocessing and model inference, thereby requiring only user-provided materials. However, it offers relatively less flexibility. An example command is as follows:
+```bash
+python vace/vace_pipeline.py --base wan --task depth --video <path-to-video> --prompt <prompt>  # wan
+python vace/vace_pipeline.py --base lxt --task depth --video <path-to-video> --prompt <prompt>  # ltx
+```
+
+### 4.2 Inference Examples
+
+We provide test examples under different tasks, enabling users to validate according to their needs. These include **task**, **sub-tasks**, **original inputs** (ori_videos and ori_images), **model inputs** (src_video, src_mask, src_ref_images, prompt), and **model outputs**.
+
+| task        | subtask            | src_video                                                                                                            | src_mask                                                                                                           | src_ref_images                                                                                                                                                                                                                                                                                                | out_video                                                                                                              | prompt                                                                                                                                                                                                                                                                                                                                            | ori_video                                                                                                            | ori_images                                                                                                                                                                                                                                                                                            |
+|-------------|--------------------|----------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+| txt2vid     | txt2vid            |                                                                                                                      |                                                                                                                    |                                                                                                                                                                                                                                                                                                               | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/txt2vid/out_video.mp4"></video>            | 狂风巨浪的大海，镜头缓缓推进，一艘渺小的帆船在汹涌的波涛中挣扎漂荡。海面上白沫翻滚，帆船时隐时现，仿佛随时可能被巨浪吞噬。天空乌云密布，雷声轰鸣，海鸥在空中盘旋尖叫。帆船上的人们紧紧抓住缆绳，努力保持平衡。画面风格写实，充满紧张和动感。近景特写，强调风浪的冲击力和帆船的摇晃                                                                |                                                                                                                      |                                                                                                                                                                                                                                                                                                       |
+| extension   | firstframe         | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/expand_anything/out_video.mp4"></video>       | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/firstframe/src_mask.mp4"></video>      |                                                                                                                                                                                                                                                                                                               | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/firstframe/out_video.mp4"></video>         | 纪实摄影风格，前景是一位中国越野爱好者坐在越野车上，手持车载电台正在进行通联。他五官清晰，表情专注，眼神坚定地望向前方。越野车停在户外，车身略显脏污，显示出经历过的艰难路况。镜头从车外缓缓拉近，最后定格在人物的面部特写上，展现出他的坚定与热情。中景到近景，动态镜头运镜。                                                                    |                                                                                                                      | <img  style="width: auto; height: 200px; object-fit: contain;" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/firstframe/ori_image_1.png">                                                                                                                                                            |
+| repainting  | inpainting         | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/inpainting/src_video.mp4"></video>       | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/inpainting/src_mask.mp4"></video>      |                                                                                                                                                                                                                                                                                                               | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/inpainting/out_video.mp4"></video>         | 一只巨大的金色凤凰从繁华的城市上空展翅飞过，羽毛如火焰般璀璨，闪烁着温暖的光辉，翅膀雄伟地展开。凤凰高昂着头，目光炯炯，轻轻扇动翅膀，散发出淡淡的光芒。下方是熙熙攘攘的市中心，人群惊叹，车水马龙，红蓝两色的霓虹灯在夜空下闪烁。镜头俯视城市街道，捕捉这一壮丽的景象，营造出既神秘又辉煌的氛围。                                                | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/inpainting/ori_video.mp4"></video>       |                                                                                                                                                                                                                                                                                                       |
+| repainting  | outpainting        | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/outpainting/src_video.mp4"></video>      | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/outpainting/src_mask.mp4"></video>     |                                                                                                                                                                                                                                                                                                               | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/outpainting/out_video.mp4"></video>        | 赛博朋克风格，无人机俯瞰视角下的现代西安城墙，镜头穿过永宁门时泛起金色涟漪，城墙砖块化作数据流重组为唐代长安城。周围的街道上流动的人群和飞驰的机械交通工具交织在一起，现代与古代的交融，城墙上的灯光闪烁，形成时空隧道的效果。全息投影技术展现历史变迁，粒子重组特效细腻逼真。大远景逐渐过渡到特写，聚焦于城门特效。                              | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/outpainting/ori_video.mp4"></video>      |                                                                                                                                                                                                                                                                                                       |
+| control     | depth              | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/depth/src_video.mp4"></video>            |                                                                                                                    |                                                                                                                                                                                                                                                                                                               | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/depth/out_video.mp4"></video>              | 一群年轻人在天空之城拍摄集体照。画面中，一对年轻情侣手牵手，轻声细语，相视而笑，周围是飞翔的彩色热气球和闪烁的星星，营造出浪漫的氛围。天空中，暖阳透过飘浮的云朵，洒下斑驳的光影。镜头以近景特写开始，随着情侣间的亲密互动，缓缓拉远。                                                                                                            | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/depth/ori_video.mp4"></video>            |                                                                                                                                                                                                                                                                                                       |
+| control     | flow               | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/flow/src_video.mp4"></video>             |                                                                                                                    |                                                                                                                                                                                                                                                                                                               | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/flow/out_video.mp4"></video>               | 纪实摄影风格，一颗鲜红的小番茄缓缓落入盛着牛奶的玻璃杯中，溅起晶莹的水花。画面以慢镜头捕捉这一瞬间，水花在空中绽放，形成美丽的弧线。玻璃杯中的牛奶纯白，番茄的鲜红与之形成鲜明对比。背景简洁，突出主体。近景特写，垂直俯视视角，展现细节之美。                                                                                                    | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/flow/ori_video.mp4"></video>             |                                                                                                                                                                                                                                                                                                       |
+| control     | gray               | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/gray/src_video.mp4"></video>             |                                                                                                                    |                                                                                                                                                                                                                                                                                                               | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/gray/out_video.mp4"></video>               | 镜头缓缓向右平移，身穿淡黄色坎肩长裙的长发女孩面对镜头露出灿烂的漏齿微笑。她的长发随风轻扬，眼神明亮而充满活力。背景是秋天红色和黄色的树叶，阳光透过树叶的缝隙洒下斑驳光影，营造出温馨自然的氛围。画面风格清新自然，仿佛夏日午后的一抹清凉。中景人像，强调自然光效和细腻的皮肤质感。                                                              | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/gray/ori_video.mp4"></video>             |                                                                                                                                                                                                                                                                                                       |
+| control     | pose               | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/pose/src_video.mp4"></video>             |                                                                                                                    |                                                                                                                                                                                                                                                                                                               | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/pose/out_video.mp4"></video>               | 在一个热带的庆祝派对上，一家人围坐在椰子树下的长桌旁。桌上摆满了异国风味的美食。长辈们愉悦地交谈，年轻人兴奋地举杯碰撞，孩子们在沙滩上欢乐奔跑。背景中是湛蓝的海洋和明亮的阳光，营造出轻松的气氛。镜头以动态中景捕捉每个开心的瞬间，温暖的阳光映照着他们幸福的面庞。                                                                              | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/pose/ori_video.mp4"></video>             |                                                                                                                                                                                                                                                                                                       |
+| control     | scribble           | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/scribble/src_video.mp4"></video>         |                                                                                                                    |                                                                                                                                                                                                                                                                                                               | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/scribble/out_video.mp4"></video>           | 画面中荧光色彩的无人机从极低空高速掠过超现实主义风格的西安古城墙，尘埃反射着阳光。镜头快速切换至城墙上的砖石特写，阳光温暖地洒落，勾勒出每一块砖块的细腻纹理。整体画质清晰华丽，运镜流畅如水。                                                                                                                                                    | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/scribble/ori_video.mp4"></video>         |                                                                                                                                                                                                                                                                                                       |
+| control     | layout             | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/layout/src_video.mp4"></video>           |                                                                                                                    |                                                                                                                                                                                                                                                                                                               | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/layout/out_video.mp4"></video>             | 视频展示了一只成鸟在树枝上的巢中喂养它的幼鸟。成鸟在喂食的过程中，幼鸟张开嘴巴等待食物。随后，成鸟飞走，幼鸟继续等待。成鸟再次飞回，带回食物喂养幼鸟。整个视频的拍摄角度固定，聚焦于巢穴和鸟类的互动，背景是模糊的绿色植被，强调了鸟类的自然行为和生态环境。                                                                                      | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/layout/ori_video.mp4"></video>           |                                                                                                                                                                                                                                                                                                       |
+| reference   | face               |                                                                                                                      |                                                                                                                    | <img  style="width: auto; height: 200px; object-fit: contain;" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/face/src_ref_image_1.png">                                                                                                                                                                      | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/face/out_video.mp4"></video>               | 视频展示了一位长着尖耳朵的老人，他有一头银白色的长发和小胡子，穿着一件色彩斑斓的长袍，内搭金色衬衫，散发出神秘与智慧的气息。背景为一个华丽宫殿的内部，金碧辉煌。灯光明亮，照亮他脸上的神采奕奕。摄像机旋转动态拍摄，捕捉老人轻松挥手的动作。                                                                                                      |                                                                                                                      | <img  style="width: auto; height: 200px; object-fit: contain;" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/face/ori_image_1.png">                                                                                                                                                                  |
+| reference   | object             |                                                                                                                      |                                                                                                                    | <img  style="width: auto; height: 200px; object-fit: contain;" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/object/src_ref_image_1.png">                                                                                                                                                                    | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/object/out_video.mp4"></video>             | 经典游戏角色马里奥在绿松石色水下世界中，四周环绕着珊瑚和各种各样的热带鱼。马里奥兴奋地向上跳起，摆出经典的欢快姿势，身穿鲜明的蓝色潜水服，红色的潜水面罩上印有“M”标志，脚上是一双潜水靴。背景中，水泡随波逐流，浮现出一个巨大而友好的海星。摄像机从水底向上快速移动，捕捉他跃出水面的瞬间，灯光明亮而流动。该场景融合了动画与幻想元素，令人惊叹。 |                                                                                                                      | <img  style="width: auto; height: 200px; object-fit: contain;" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/object/ori_image_1.png">                                                                                                                                                                |
+| composition | reference_anything |                                                                                                                      |                                                                                                                    | <img  style="width: auto; height: 200px; object-fit: contain;" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/reference_anything/src_ref_image_1.png">,<img  style="width: auto; height: 200px; object-fit: contain;" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/reference_anything/src_ref_image_2.png"> | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/reference_anything/out_video.mp4"></video> | 一名打扮成超人的男子自信地站着，面对镜头，肩头有一只充满活力的毛绒黄色鸭子。他留着整齐的短发和浅色胡须，鸭子有橙色的喙和脚，它的翅膀稍微展开，脚分开以保持稳定。他的表情严肃而坚定。他穿着标志性的蓝红超人服装，胸前有黄色“S”标志。斗篷在他身后飘逸。背景有行人。相机位于视线水平，捕捉角色的整个上半身。灯光均匀明亮。                           |                                                                                                                      | <img  style="width: auto; height: 200px; object-fit: contain;" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/reference_anything/ori_image_1.png">,<img  style="width: auto; height: 200px; object-fit: contain;" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/reference_anything/ori_image_2.png"> |
+| composition | swap_anything      | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/swap_anything/src_video.mp4"></video>    | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/swap_anything/src_mask.mp4"></video>   | <img  style="width: auto; height: 200px; object-fit: contain;" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/swap_anything/src_ref_image_1.png">                                                                                                                                                             | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/swap_anything/out_video.mp4"></video>      | 视频展示了一个人在宽阔的草原上骑马。他有淡紫色长发，穿着传统服饰白上衣黑裤子，动画建模画风，看起来像是在进行某种户外活动或者是在进行某种表演。背景是壮观的山脉和多云的天空，给人一种宁静而广阔的感觉。整个视频的拍摄角度是固定的，重点展示了骑手和他的马。                                                                                        | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/swap_anything/ori_video.mp4"></video>    | <img  style="width: auto; height: 200px; object-fit: contain;" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/swap_anything/ori_image_1.jpg">                                                                                                                                                         |
+| composition | expand_anything    | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/expand_anything/src_video.mp4"></video>  | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/expand_anything/src_mask.mp4"></video> | <img  style="width: auto; height: 200px; object-fit: contain;" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/expand_anything/src_ref_image_1.png">                                                                                                                                                           | <video controls height="200" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/expand_anything/out_video.mp4"></video>    | 古典油画风格，背景是一条河边，画面中央一位成熟优雅的女人，穿着长裙坐在椅子上。她双手从怀里取出打开的红色心形墨镜戴上。固定机位。                                                                                                                                                                                                                  |                                                                                                                      | <img  style="width: auto; height: 200px; object-fit: contain;" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/expand_anything/ori_image_1.jpeg">,<img  style="width: auto; height: 200px; object-fit: contain;" src="https://huggingface.co/datasets/ali-vilab/VACE-Benchmark/resolve/main/assets/examples/expand_anything/ori_image_2.png">      |
+
+## 5. Limitations
+
+- VACE-LTX-Video-0.9
+  - The prompt significantly impacts video generation quality on LTX-Video. It must be extended in accordance with the methods described in this [system prompt](https://huggingface.co/spaces/Lightricks/LTX-Video-Playground/blob/main/assets/system_prompt_i2v.txt). We also provide input parameters for using prompt extension (--use_prompt_extend).
+  - This model is intended for experimental research validation within the VACE paper and may not guarantee performance in real-world scenarios. However, its inference speed is very fast, capable of creating a video in 25 seconds with 40 steps on an A100 GPU, making it suitable for preliminary data and creative validation.
+- VACE-Wan2.1-1.3B-Preview
+  - This model mainly keeps the original Wan2.1-T2V-1.3B's video quality while supporting various tasks. 
+  - When you encounter failure cases with specific tasks, we recommend trying again with a different seed and adjusting the prompt.
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/depth/ori_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/depth/ori_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..fbbbbe882bc3f40366153818625cf7bd2663b353
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/depth/ori_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a155ff5265ac0f4beca76257744fff89182ed1243caeb31fe50eba4c442ac561
+size 9175298
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/depth/out_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/depth/out_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..dacf8bae22177cae9d8ba142c4a2d2b97d293e95
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/depth/out_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5bb666d8ca82ce9f0815a666639d6a71b9e716b1c8283d1f5e3f266c7e3b1567
+size 3844987
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/depth/src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/depth/src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..ddfffb06172f1e8b0605ab0ed3e9d899689e4b16
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/depth/src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f201fd22fb40b56e43093cd8d22851d67b4fd2186c35cf24b9dd4ab95ca3ced0
+size 2673490
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/expand_anything/ori_image_1.jpeg b/video_gen_14d/data/VACE-Benchmark/assets/examples/expand_anything/ori_image_1.jpeg
new file mode 100644
index 0000000000000000000000000000000000000000..e83e4c031c0c2bd7cb17640578212a4f0ca40e22
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/expand_anything/ori_image_1.jpeg
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e4d6c64c0d8d7cef0b70a9cca87f4d20b65d4c7653ccac2ccb0b6ec50b68ec19
+size 334372
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/expand_anything/ori_image_2.png b/video_gen_14d/data/VACE-Benchmark/assets/examples/expand_anything/ori_image_2.png
new file mode 100644
index 0000000000000000000000000000000000000000..da8035b7faad464ca3d0976d843f9c39e41a0b3a
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/expand_anything/ori_image_2.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f68ffd0de6498a184a0a27d66214d0fde3292878f2f3c6d28c736fb37cf030e
+size 505308
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/expand_anything/out_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/expand_anything/out_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..9955603f1f59c560beef3529a0d045f644c1d9b5
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/expand_anything/out_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f0e0669ce7ecd4298b10e1d6dfd16289a51ea70b41071905493a8063d0914362
+size 5539047
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/expand_anything/src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/expand_anything/src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..da418bbfa79b75f41451c2542f18118e27feb2f6
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/expand_anything/src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc1a626af01726336d240c564b111bb7d3bc7afab7a507e449745b83feb2a1f9
+size 6127
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/expand_anything/src_ref_image_1.png b/video_gen_14d/data/VACE-Benchmark/assets/examples/expand_anything/src_ref_image_1.png
new file mode 100644
index 0000000000000000000000000000000000000000..da8035b7faad464ca3d0976d843f9c39e41a0b3a
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/expand_anything/src_ref_image_1.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f68ffd0de6498a184a0a27d66214d0fde3292878f2f3c6d28c736fb37cf030e
+size 505308
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/expand_anything/src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/expand_anything/src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..fa5d54c2b2e4e90ebba7f72aa377bc5050895bab
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/expand_anything/src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5552f0564a8f59e61e7b974e9129255a73a086e5b1ccb718d1aabf60b42e8cd
+size 176778
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/face/ori_image_1.png b/video_gen_14d/data/VACE-Benchmark/assets/examples/face/ori_image_1.png
new file mode 100644
index 0000000000000000000000000000000000000000..1af5854eb47dda92d9d4093b0cf26ded58e71c05
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/face/ori_image_1.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0c49bdee86644a0e8c81a719ead4d5fa2d82a4150c44387c24db05bd124ad655
+size 1027739
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/face/out_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/face/out_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..c4d1b0dd8b5f0197e9ac85c498667b3b60323647
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/face/out_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb47e38d2487e05c5f25e3182e11373164c905d11562a5a67feb0201f486c3b2
+size 4738703
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/face/src_ref_image_1.png b/video_gen_14d/data/VACE-Benchmark/assets/examples/face/src_ref_image_1.png
new file mode 100644
index 0000000000000000000000000000000000000000..643e0a9838df53d160e90e4ef2084832ecc1d66e
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/face/src_ref_image_1.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6bd58e69f1a39b7d587fb30e651d1b9121f66d72969e4ad498a2eeaa7287e94
+size 294376
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/firstframe/ori_image_1.png b/video_gen_14d/data/VACE-Benchmark/assets/examples/firstframe/ori_image_1.png
new file mode 100644
index 0000000000000000000000000000000000000000..777bae1b55d663c9ad4e76ca84da5bee445ec2dd
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/firstframe/ori_image_1.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f349c25338856941ac115294526f61e0597614c89820cc667cfe8288e2e995f6
+size 1176227
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/firstframe/out_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/firstframe/out_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..ee5ef0a417f97b5d49c1f3aae7de0ad4e177afd3
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/firstframe/out_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d2f4ef6fb6aad1fdb25437f3c12d0739c705d0b5c61cedcb4b17888e1ffa3f0
+size 3184400
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/firstframe/src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/firstframe/src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..eb98757846f7388d5a79c888880a6fb2a0c6aab3
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/firstframe/src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1cfc44fb17decc101b8f9a02483b61c66dbb205816559e9b0841e1c361e2e72e
+size 10356
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/firstframe/src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/firstframe/src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..f50e0650a8e2fb99149bdf28f83ff1c1d3a3ff70
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/firstframe/src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:543b372988f337725881323671a44697d082be3ffb77059edd13d83e9b0511dd
+size 152102
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/flow/ori_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/flow/ori_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..92e1bd8e67b8edbeec07ca4ca2117d03c901c701
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/flow/ori_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:649411395d1aee25aa5c65a61ca904eabfda2d56a920923db1abac5fc4942b4a
+size 7033492
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/flow/out_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/flow/out_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..1484dc5358f6970736bd340a8cc40755ffcc230a
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/flow/out_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f66c40fa0f9d8b8aff3c82e6e20264bbee29ccf0d6544662871c6f46d26d349
+size 2215940
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/flow/src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/flow/src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..bad25edfd2723ab8ce5ae2485ec69f5a936b019a
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/flow/src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:992309a0bd692b870afb601400d26638b0db2d84eeafa1d466b30a7739d02bdd
+size 5484810
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/gray/ori_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/gray/ori_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..1c9eba7af83642f1f33616221729d59b3bafdf5f
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/gray/ori_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:daa4468a7698a8c5901bdd919bafc9c707e1e9d58fd248c1f40a0107ce1d1287
+size 9176438
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/gray/out_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/gray/out_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..b66fceed3186c246eea8d79bfbb66230ee5044a4
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/gray/out_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89c40312a3be1d48be1ecce48069d66077a438afda32417884e5952882bb522a
+size 3823543
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/gray/src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/gray/src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..0bdd728271bb98dea0f645dbd9fb3d0ab2e7fa43
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/gray/src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8060f5e29a4623c0417e45c241a1f06f9004eb36a6e871da3827cb949257f32e
+size 8583598
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/inpainting/ori_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/inpainting/ori_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..1076845758fd8e6d7352229c35c80a897c9be438
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/inpainting/ori_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6ff125a102beae9305ba0e04c536ad5e401c82a46b2b53a78df58ef174181b6
+size 14418789
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/inpainting/out_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/inpainting/out_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..f64bcaf3558f0e8e4a94f1c5f0ef3cd4c8db7d31
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/inpainting/out_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b532ed0fc4da14d81e0837c946b294ea5b1ef2627aa4299e57c1324d2f6339c5
+size 6472206
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/inpainting/src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/inpainting/src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..79696c6baa47b845bd3421ea9a11963050f81804
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/inpainting/src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:07e7f67f6389eb5172c4531a3d9fe73c6893e47fedd00f973d543bb06520d851
+size 378282
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/inpainting/src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/inpainting/src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..0310310a55a85836f86319fd64be94994e446efd
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/inpainting/src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6de96dd575c79437184710a847929b4bc06501ebc6a48ba34631707e4367e29
+size 7579044
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/layout/ori_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/layout/ori_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..4c89452203bd445e98244a41aae507cc594e764b
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/layout/ori_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d5934f2695ab878724d129961047a28f8641b19abf5f9a0d6f1e555046786c1
+size 702795
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/layout/out_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/layout/out_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..96af6820fc32d133e63a8b6599a919f185dc2f3a
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/layout/out_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f25619eda00e56b34018b4418c9924441d44fef6cde1602a5476773e5ead5eb8
+size 4480934
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/layout/src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/layout/src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..fdef23558f12f67b7c6b06971aa36210e7fff723
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/layout/src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3a67e02372bef015fc657a13d5aa490e758d58114d237978250af7f96293628
+size 20003
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/object/ori_image_1.png b/video_gen_14d/data/VACE-Benchmark/assets/examples/object/ori_image_1.png
new file mode 100644
index 0000000000000000000000000000000000000000..eaba2fcaf52f03b39839ddbff2f2c2feb4538675
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/object/ori_image_1.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b28aaf2acbda34a6c13133be6abf8478fac4693f27b382b361028a202a06894
+size 1183904
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/object/out_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/object/out_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..03dba1421a1ad31b04b02e12fd94f300ea92c953
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/object/out_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc5b17704605550901c5e4f33292136fb6dfd00a7d27686f50dfd5cb356b1bdc
+size 5980751
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/object/src_ref_image_1.png b/video_gen_14d/data/VACE-Benchmark/assets/examples/object/src_ref_image_1.png
new file mode 100644
index 0000000000000000000000000000000000000000..e88b9ee138a9c8a7264c2724fa32d5bb56fdd31b
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/object/src_ref_image_1.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1cf113c9df030a5547ec04163bd0490e81da4853445955bb752f50570a48ccee
+size 514750
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/outpainting/ori_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/outpainting/ori_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..9c1ca3daeda2996fb6aad1e382980411264a75bf
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/outpainting/ori_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3be20246b73a8ddc376b54b20628e87e5b68b8d534774d4cf9635ac99673a856
+size 20777667
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/outpainting/out_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/outpainting/out_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..8abdd3b48632c71767a82543134bd320a6a9d0aa
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/outpainting/out_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3753f89aaf5fce98d656ac100d93f20fc65b31750c088f99c2fb9238de7ab0e
+size 7588273
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/outpainting/src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/outpainting/src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..11ea534b361311f1c969d050ba3a5de5184d46c1
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/outpainting/src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b2bbb16a03d84c0ff3c78d543fd2630f652315bff4e8c28fed7856a7f97110cc
+size 16731
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/outpainting/src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/outpainting/src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..6179e87a15c0fb13b333147ee3c7c8df16781285
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/outpainting/src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a797263380421b892cd4d43dedc9e3652d5c345de68c4d2f072e57a733cc8cf1
+size 20695622
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/pose/ori_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/pose/ori_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..67fda5d6b34cb49f7bfa3e6bc125a5f676f0be56
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/pose/ori_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:005e850eb0fc27ccade3967e143df7155754868199290549b9e9eb4ec298a558
+size 7007490
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/pose/out_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/pose/out_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..1facc15a4f17bbb953e150f36d5dae699d68ac4f
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/pose/out_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ad90520040dafd9cddc5ac915a61cdce7394105e1d56cfbb9cf8363f2c85086
+size 5844282
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/pose/src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/pose/src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..fd3e534f40a39293a88dfcb36ce6ba900b6bc5f0
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/pose/src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c6c8fe70de71f0a6768cb46cce3c22d1e9934926a07e6e3993bc4416b71bf6d
+size 3431100
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/reference_anything/ori_image_1.png b/video_gen_14d/data/VACE-Benchmark/assets/examples/reference_anything/ori_image_1.png
new file mode 100644
index 0000000000000000000000000000000000000000..49efdbe961b74e21cc013205b6934361f17cd7d9
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/reference_anything/ori_image_1.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5b8d183558c291a9257f744260e69644392fe920663ab8ff56269f63cf58900
+size 1922395
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/reference_anything/ori_image_2.png b/video_gen_14d/data/VACE-Benchmark/assets/examples/reference_anything/ori_image_2.png
new file mode 100644
index 0000000000000000000000000000000000000000..d975d64065d8f14ab80a582f04ceab6b39f425ba
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/reference_anything/ori_image_2.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68af093331dd760b550d7d247f0d5e341524d7602919585085a3ed75224e7e09
+size 422455
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/reference_anything/out_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/reference_anything/out_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..d179060e9bdcc5ba76d07a2ce20c212dbeea2825
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/reference_anything/out_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc0e1db04df73332af56cef84b86b35e7dc80df3be08292e7064b3db40c91692
+size 3530325
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/reference_anything/src_ref_image_1.png b/video_gen_14d/data/VACE-Benchmark/assets/examples/reference_anything/src_ref_image_1.png
new file mode 100644
index 0000000000000000000000000000000000000000..4a395e90472be7d6343bd88be388a26382f59d67
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/reference_anything/src_ref_image_1.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7786b2ac0ab20a8e988429fb99758930a3a66d1e309c5656eef9b8042e5f2922
+size 407945
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/reference_anything/src_ref_image_2.png b/video_gen_14d/data/VACE-Benchmark/assets/examples/reference_anything/src_ref_image_2.png
new file mode 100644
index 0000000000000000000000000000000000000000..ac29d60a35cde57bf0c8e28d6bcdd1496ae3fade
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/reference_anything/src_ref_image_2.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8ec7050feff4dc993df42ca158522c5c6000a0779c19e815ffa57d1b49eaae1
+size 383921
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/scribble/ori_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/scribble/ori_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..c4eac39b83954a5c4dd0c20606c28692c7d807b3
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/scribble/ori_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:030f88eb7f4a8fa0ea927be8525609c995acf90fb6b0bdfdc3b5da1c695955e0
+size 14542584
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/scribble/out_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/scribble/out_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..ae5b6b509ec64cd9503bfa56450a7decf627f148
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/scribble/out_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:786e2900ab5b93536254d10b1d768dfe99c1caba2d29a94906cf1d6def101a19
+size 5798493
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/scribble/src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/scribble/src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..357ed4dd2b254dc539fb1b78eaafefab8b1b8e53
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/scribble/src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0caaedd0c3126979471418b408acd7653f77610eccf086e4a197a7277f79358c
+size 16876466
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/swap_anything/ori_image_1.jpg b/video_gen_14d/data/VACE-Benchmark/assets/examples/swap_anything/ori_image_1.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..c83bb22f54b831e0838487ad796a33f3f0234f3e
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/swap_anything/ori_image_1.jpg
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64021ea3e60a94a159eb2a34998904e47d47c5314f357892b9a29fa0018be6c9
+size 73264
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/swap_anything/ori_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/swap_anything/ori_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..6b6cc810cfd60f119a25889dfb51357986b7ed15
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/swap_anything/ori_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a1b04900f989e177841163356835f6f4e036b86bffe2709d4918b8fe7562a63
+size 2537893
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/swap_anything/out_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/swap_anything/out_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..19deb5be04e6fabd3d7dfe7781667880109dfe9b
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/swap_anything/out_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8cd10032517bc363b84915f50e5b3a815b807416784de95a28dcb9720730745e
+size 2554786
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/swap_anything/src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/swap_anything/src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..3d2a17245c222335a8e329d24bc35a29ae559055
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/swap_anything/src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b0ddaac1eb44583948f87249b7fe20541eaafd375c5499eb7cf7831b889bb2f
+size 9700
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/swap_anything/src_ref_image_1.png b/video_gen_14d/data/VACE-Benchmark/assets/examples/swap_anything/src_ref_image_1.png
new file mode 100644
index 0000000000000000000000000000000000000000..ab43c6a4dddafaf88daa6bd1c74fb5a3172f9c65
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/swap_anything/src_ref_image_1.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3b3b56a11e9f0c33a1c406fda6c254f908e176dd05a079f77d6b7a8d9a4dd3b
+size 178518
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/swap_anything/src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/swap_anything/src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..aa4bc29ff7966881564248f03bbca58407c58c00
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/swap_anything/src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f722afa11cdaf7b183ab6d50cb77ee29b60a131ac6ba475b0199cadb12966fbf
+size 342837
diff --git a/video_gen_14d/data/VACE-Benchmark/assets/examples/txt2vid/out_video.mp4 b/video_gen_14d/data/VACE-Benchmark/assets/examples/txt2vid/out_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..b9db39cacc76d9fb2dabc9298cfdea1d568f8313
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/assets/examples/txt2vid/out_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d417a467056872aa5834c79d49d1c9aa8b112a3d24a7ad438a26f0aecaf7d61
+size 5361588
diff --git a/video_gen_14d/data/VACE-Benchmark/config.json b/video_gen_14d/data/VACE-Benchmark/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000000_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000000_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..f68817c0692dcef8ed106cef9b5a66878ee7b7a5
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000000_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc9410ed6ad9a32cddfcf51f35e2ef8539ff294767544f9109fb12ba0dfa2f7f
+size 1231563
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000000_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000000_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..19b71e28ee41e5e54c1c8e8bae7a58ec35d2d604
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000000_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:41752d0bae5a9d9a802b8e85deab61d9fbbe59e37254914967dea047dcb8277b
+size 15259
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000000_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000000_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..99403dfe5bd5c0320d3b6ce2a2d6d63758ac3f8f
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000000_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d75825c20cd1b8fa41b1b005d541759373381830966d5d91755f86d191e89dbd
+size 15256
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000001_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000001_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..104205cc6ac31845c0b45379fe0b766708a3d95b
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000001_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3599be7932ad12c7e7b1435bd1406bb1242a17b139d24e380bea54c6fe663d6
+size 6149441
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000001_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000001_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..fa9ce10f016befc0bf848b59a47b858cf6c464a8
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000001_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a2bcea7d058f773d8fe995aeb65d963ce31bad5e65b44585098f5cc273f1902a
+size 9344
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000001_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000001_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..428ea02b73c763f5ac7c79dcb07da738c1f6d5d5
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000001_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:adccb70bbdca8d77cc18820d3954f9f2e2981b9202591732b1cfadd75ebf4d8c
+size 9342
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000002_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000002_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..9902e9dc979d317ca2caec2bc757648c6bb947fb
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000002_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e62fe6c0ff5750770b1e9569f7f7c637bfcacd15e4d28e699484a7d60b7457e
+size 1485085
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000002_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000002_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..9071e3849d854730084211df091a650db8b6766f
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000002_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d738e3dc8afb2dd0f1da94c254cd62921ef8286376e96fd1b1b2a33470a2e71
+size 6919
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000002_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000002_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..3f3768dfc76708e3ece451b1337b467e8b8dff0e
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000002_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f369fa47ac7ab423788115ae49ed693d775939233149de50a3ed545a24cd466
+size 6917
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000003_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000003_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..70d3ae8870fe8deb898993dfd3fe204ccbfd7d00
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000003_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31d8c8163f77e9c0b2c6fe74c50d2df3b778104414a6c66edebb6203d1be0a2b
+size 413732
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000003_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000003_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..16d27bffa065354db4b8984cb6b12b3366c1e52b
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000003_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:11bcf90e1cb52621a8d1582f1ffac28a193569bd437a07d6dfea6dbcdb1a1529
+size 13539
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000003_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000003_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..785d6fa8d07fcf57764f1a56b5edd1711dd39083
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000003_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68d909092cce71a88d914f299c2c53a07329dd30241855be5af492a9eeb38146
+size 13538
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000004_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000004_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..c8d6f5f9d566a804514473aa8013abd538df7a00
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000004_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c23530886d6fec5e640a87dbb34c02e2a8e5ee6949d18c1c98acbbab872f75d
+size 264348
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000004_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000004_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..da69a16f54b8f0e81249de90c917f747d47c457b
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000004_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4693b4a8f850b0080d877b7abe2a215e2f93cae5c0f381531e1bbc5f74c810b
+size 6862
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000004_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000004_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..ef4fc151807bd7ea16cbe35535a9cb40c11f51d1
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000004_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9eb47d2c17d187189f5f6f4f1a3e363291993978e31547d48d15523b9d58aca2
+size 6860
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000005_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000005_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..20691541eacf598133c766744db1c7e8eab3c5b6
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000005_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bfdbc6dccfb25b870516c524b2ed0ffd8e9b9d1f12099073af6778679addcf4a
+size 233961
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000005_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000005_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..3f68760f02612c1209a3efc3839e41d6e635a83c
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000005_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:54f538bdfe1318e1cf1fb768bde446db872ebfe4b7c2e2918638f989fc639f6d
+size 6973
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000005_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000005_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..616b0c6329f4453679e2889671d4377db713cca7
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000005_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c189d17a3644579412a6b2ca3e8ed883adb2f02fcff50391ecc68fe942831f7
+size 6972
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000006_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000006_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..43b6848cd653f4210ff26d83cdd0b717ce958995
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000006_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b1ce4ead4444198e9703a650b2a770b402cedd97d0ff2fdc4404d3673d84cff
+size 5607328
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000006_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000006_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..b76047f19569e2ae89571642ef3c4bdb95fa0ac7
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000006_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fcf509e182c822ab13d5bb10441daa294d699a97c2a4fe6ba02acb5aa9c4b6b3
+size 8102
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000006_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000006_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..f3226cb4e77d7ad52e687d38e245cb18b8b40b1b
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000006_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5449ffecb3c25752c2a9d458e47853f8dfd2743caac3865db15dce8fedb6413
+size 8100
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000007_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000007_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..1fa88e2843d70aa665769b5e18286ffd32e3434b
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000007_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:17622a2d682b8d66ad7a1b7fe2de18c27e4981dcb1ed92e47826d6c74efd3909
+size 339778
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000007_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000007_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..4ce145ff85a0ca1512f2b89e7f935fbca86cd319
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000007_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71fa2dc6ac8934193ad33dc883a50e60e682334c76d528e04ac954b4c8fa7a07
+size 7441
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000007_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000007_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..be0318cbc842d137bd0cf3c8ffdc1dea3e9cad87
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000007_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f863d9c1b6df432d291f03641fb305f74ddd8d62029a2b4ab8531a3bf869bf68
+size 7439
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000008_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000008_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..1649b081eaa4dc55cc7ef8de9c17b731dea070a1
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000008_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e2ed8352989eed418b1e3575e6a6722eeed332112bbeacab7a97f4a3aa85a5b
+size 979799
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000008_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000008_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..2bf4177489f90561248e29cc5bb20e4b56ff65a0
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000008_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4521eb525248e03c21e21121040e560f750c53249e2502b6cc0c8fd59e769d01
+size 14061
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000008_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000008_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..96fbb98dd84b7752c989ef0f10e470263db8fedf
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000008_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2426bc1d17f757166a988432669ea1f96bba6bdad2b329934d3205ad2d56e451
+size 14060
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000009_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000009_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..1e824f03624df92eb14575b927f8c13a6686d740
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000009_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9fa28ffaed166f0cc95ea67e65119c035cff338aa1489d6f4cc29d85ca580aa2
+size 3828991
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000009_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000009_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..1f0d906468053916b0cc9aeacc6455ae818845b9
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000009_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3493d15d12555ebe9d9391486439ae18337231de74314efbefd4c4a1e884ff59
+size 15159
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000009_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000009_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..3a9f555d9a4c9407b32920bdce9b0c438dd49756
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000009_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:38f13700351c865c904aba11c15f9f102a792b94bbca9b6cd38af1d8942f23f6
+size 15155
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000010_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000010_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..ba7b783893221bbe4162a0fba991a62979fdd71d
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000010_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a7607bb3bec372b4e54a9c05a4e94455604233ac6807763e73c4dc19d41810a
+size 882267
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000010_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000010_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..61bad1b2d55e21996d6aa9dd414fd151600005f6
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000010_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ed67a436e073d210f79dbf820746b0742e35924895669bddd9cdec8a1f0b9fa
+size 372349
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000010_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000010_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..1fae1d7d2e2ab5879d61dfaee4cfc93deec33e4a
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000010_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df2e0d8c78cb09d623421001d888eaaa0f859ac1c37bf76efa0ab724756d46cf
+size 458678
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000011_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000011_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..af5e6b076535148fb722c9d3c6e27eec62a4173f
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000011_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a1e0591ad62b4cc161d05d71d46ac9d50566d5f448396826e49eec08dc6e5e8
+size 1029640
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000011_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000011_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..5ab5c3e050b41b8c1c9e2248ba234e2b42daecc4
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000011_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3cba7ab123a3cdad680acc670d2392108e7f7d834dc798de34f9dc177451022b
+size 14995
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000011_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000011_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..589b1a7c7cde061054c07b5d0d61b07842f9c203
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000011_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b501d29a604a64a5e7b319fd895966ca9c14e3a59900ef181be949c7cb33f4f9
+size 286155
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000012_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000012_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..09dc0e776cb87cda9c5e456fedd28389a84fb781
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000012_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6ae8b433859f44539a170116f3704e19a53fdab7385732e395cf7f297beaa38e
+size 331822
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000012_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000012_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..237a2a7638136592bf4d984f74b5692a4d2c3b88
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000012_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82e2cdca01a31d816c6b3edb15b1a7798ca776355f08bf01ef08079bfc467ffb
+size 18620
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000012_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000012_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..d59cf0fbf18d4749c39f2d51e10633100a426500
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000012_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5927008efa92be0c179ae0990d366c49f5a3ce8177018b23f73bdb1ca59ce105
+size 283535
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000013_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000013_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..5587958d81a58b883725cae149f2115fde3f057a
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000013_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a5347e29bb5ac880e5672dce30a3be780cef97679ed81153d7bba80582478e86
+size 779411
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000013_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000013_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..6defa78a89290e74ad7ee14355823b3df6a908e5
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000013_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a9301ae60f1c66dc9a512c423a8b2b8b4a57b72ce6b3badc53deaf01f92c719
+size 59481
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000013_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000013_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..792535f1e73736890a6f760e37a2ca7e02296ad6
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000013_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82a5db6a8e659300b9bd50eb684c48d48366a178e00a5f1117f2ad29090214ba
+size 512151
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000014_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000014_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..404b8bcaa3c2bc78b54aaf65ed04d48c4fb7027d
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000014_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5bc4ff6a441b69e385ce179f859184159560b918a0b8a2e51d2a492d6c2a0a55
+size 643930
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000014_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000014_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..7208c64b03ece5d1a971d0c4968fce893dea502a
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000014_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c4a5e19ff155448eefd061ea1882bbf7e88b1f8273ac8161dab8751ce8ed1d4
+size 275995
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000014_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000014_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..c4accd500dc0fbc3a37860203ed9e7dcb3b34b18
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000014_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67f2575f7f6cab570c2c7be9d2b649a990ea62f817892a3b07ad9a97cacfe649
+size 316931
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000015_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000015_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..00abec1684a2a6dc294d90229758982dcfec33d3
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000015_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b29d37ad7c003959a49004e8e3241e67b56610eb1a797a2b6c40b728caafc5f9
+size 1286877
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000015_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000015_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..d617c20a5057d41e7421381ab487861e61b2c84b
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000015_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25cc720adab78447928bf7899890ee53841f9895c6fdb0c536627841f64cc3e2
+size 102221
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000015_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000015_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..2970c83a24b6bb25992fde8cd6ab8a839862c62b
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000015_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a2a432ccebfbbd8ac2f5525ce835f13951adba076a700f9c722d6527a985d19
+size 1071147
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000016_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000016_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..d84316eb1272d021dad559bdbaaf72f711b7f16f
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000016_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18e48bfbd2c504e1e18b72b2e1f3f97bea3e1077f38b906d8e97225cee79aebd
+size 1359723
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000016_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000016_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..cc81a18503165a0272a09f1eed6eb1f37e8f0f55
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000016_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a0e6f9e99d89fd5a421e8df8096e452e10d5017ef74c7373545724f3bb76c7d
+size 89622
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000016_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000016_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..a8b71d5eeab9aed1d339d5da43c22588cd144943
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000016_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:598899a4845b53835a67ffb337cbf02ead06c5760d2ac02d62b861b4e4c52225
+size 1075391
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000017_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000017_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..470c66dd23f282759ec475f5c5a8f48c2c84ce76
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000017_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ab850d6ec7e1359bb809a0211264a9b328e8e424f228dd21835887efb61b833
+size 429473
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000017_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000017_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..fa8fbd498742013928a75099b473eb403a61b611
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000017_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d289574358cfe8947d0d14b25d1d472b9f96fa92a04984f1af2cd2fec7aab47f
+size 11329
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000017_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000017_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..6f5eafcd5658a08bc56f8abb1f59f494068b6f60
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000017_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:30187ac5a91ecad7a2c4a4ea37565abdcdb38192968d9d6dbc1e56414bed38b8
+size 308856
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000018_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000018_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..3de62f1bb7ab640da219e8f93b9d536c783c89cb
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000018_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7a4b51bf71b7cee22024a033745b3cc09a07f36fac21b2ebe43c26692ff346c5
+size 2306118
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000018_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000018_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..6180e1ef4f8c3e111933b52d53bbdf2bc3e5b717
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000018_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce7a4e1c78fc0026299831b1ceea3d3c436ce35d623a63581ce26ac3e756781d
+size 183903
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000018_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000018_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..34d5a82c3c4752f31d52722c8026427d32346297
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000018_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a51872bb5dfa14ad0207a2294f5316dae7b5de085cfa44aec04e08012dac8c2
+size 2053479
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000019_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000019_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..6bc29f1d299f51cfa533f50d57036873c68ec000
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000019_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c09a98b64114b708ac9af3b0a04efb6ee9e6c981c19a6c8a212d7051dbccd2a2
+size 323826
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000019_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000019_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..ee5ed1a72c1a9dbb5aa0d1a114034fdddd05e617
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000019_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5824d44fa2d984b9a627bfea6ad4283dcedb8f87a7ef308779834a6fc92bd49a
+size 13180
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000019_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000019_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..87a3e1b7bab5704a0ef1769d1173558d8f6f5fab
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000019_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:474206bc713dc1a3e660879eb064a89cbbf37233c66c054df53ef6f1202aa678
+size 54891
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000020_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000020_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..b25765cf8db84aa13effc09bc71499cda88fc03e
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000020_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0e8f9a54fdd47bf667cc28e9d939b1edce4ae8dc6b0c75356ce32c032dba56fe
+size 1154680
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000020_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000020_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..98db44bee5f4628dafa03cd0d787ca01f4212f40
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000020_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9160f0e471250cb55d6c1dff41e8f7ecdded35575831d2ac86f702a5f8423383
+size 15824
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000020_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000020_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..f687bc4f1bb00524309d6bbbe6a312a10914ce16
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000020_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4dd55f270ab842c50dec91a4d6db644a4b315bb2f59c86973d9d12f95946f88
+size 4406053
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000021_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000021_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..f72ba832909f83ebf90df075a29b3ab170db23b6
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000021_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e727c6321f1a920e7cd46ce3907b01894a29c6ffd14a7896935afb62162a3056
+size 770337
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000021_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000021_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..6c6f3f3de5ea8d12ca970d5f768bc717720dc661
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000021_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5274a7d8273c1fe84a83dc0a6050a578b6cc954a0582b3202d8c4f96d273e50c
+size 8653
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000021_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000021_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..97cc28a6e64d76d8164a70debacaddd2c2666d76
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000021_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1c33ba67744e4a4dada1c7269b58d5f2e0a693187a4f50b2b8bdabd26465664
+size 3678752
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000022_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000022_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..684b957015ba4881b8535545d6a9e2e4dec41d4d
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000022_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f27adc58c6037d8c8c1f7c92e213306998411f151d448d656b49357fd063992d
+size 1366192
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000022_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000022_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..91f48570ba3bb2acb90570dd4d8bebad38d18e7a
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000022_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b11a4bc85a7d029c5e8510b6e9897160f824620bbbc2313d5a7d720fd1370222
+size 9348
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000022_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000022_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..5afd799144bf1399b1a6c76636fb50f292789153
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000022_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f532e2cf7916a9d090ec5e718b92ccc31caee53a43e184220676e84dfee833a7
+size 4683436
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000023_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000023_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..3fbfc474171a25a7cd8ae9188fbcc61a268cd9ab
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000023_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b8498b67f6426bcff7e95f9e98f31c489cf30caef09b0b570189ab3924951e7f
+size 1168792
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000023_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000023_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..dba2a3dfc5a0e6c709258bb95e8946e5b1f1e708
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000023_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1331027ef271439a641e349e7933d210a9231586d2f21e6d12da9158729acded
+size 10912
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000023_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000023_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..cdefe8f9979c4eddd594003a7a4ef5f2f2f0007e
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000023_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b33aac21d40fcfe360cc77b26ecd15707f84e72ea6eb373547ea03dc89388e6
+size 4651207
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000024_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000024_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..78aaa96ae3de244b8e7683991ce123b7d161f527
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000024_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:580036bc500e80731e2f18363cc813d0ba92b96f10afb16d85bc8a3eabe5b521
+size 1191339
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000024_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000024_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..fedcf06028524583b65ce76161f58012384e0d7d
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000024_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b5bbc3f6332ba74966b27ed71cb53823de81508c288427b1968c34154d0df31
+size 10442
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000024_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000024_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..230b120ef8c227c245e6bfdac8f81ff135b52b9e
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000024_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ba1c80ad70e1525f9b72393b3ed62066dd05bc32e57c25f1ea43b02b11dd575
+size 5381429
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000025_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000025_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..7b61e5eea725a331b652c88176a9e0e84dd21b8d
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000025_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c2821d968f3ad61f50437dc4300b6490bba1b0995e580bd923cfffcf66154304
+size 663478
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000025_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000025_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..ce4e3b60684e66fe872c21eed57f212c237ba1bd
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000025_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:41038b934f422dc62ee5b20ebeb582f6bd259869192d2eec4ea823d459ed2182
+size 16783
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000025_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000025_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..7f0e723bd8ae1d0474d4b706185654b1115acb8b
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000025_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6529c6b5de1915f3f2a35edc56da94f87a6d7133e2e27edf90a4f75dc04a4d40
+size 2819359
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000026_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000026_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..dea11b2a7b104b4a1d36ab3ba836c2ee61c799e3
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000026_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc418519432902f5be700beada6e91958cc7307da8fa2ee8aadcde4f96422758
+size 1374651
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000026_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000026_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..ad07be6ff387bc9fca903b0bf3f41f4ad642203f
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000026_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56593ce3c43a35d1915e81828f14e2ccbdb5135661b3d00e3be82e95dd833c7d
+size 8788
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000026_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000026_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..c26b4ebf5a092218afc07acddedbf561982adb3f
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000026_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:083043531c65b13be8a519f2d650db56c74ee090b8d6c544e927729509193ab6
+size 5031716
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000027_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000027_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..a6c989b1e96013aabf828d6f7d638de066932264
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000027_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc23e9a07396d889770a74fa1e1f8afe938365083b74094bfc94f88e6d7c8369
+size 352671
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000027_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000027_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..7e4779bcc0471bcaa1bce60151450e287869914d
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000027_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3daa74d29ef604955de071af1edfb1ced851b6eb34b87f5eb131dfa2494897f
+size 8170
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000027_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000027_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..4a0cbf0a0471aea408de1d1be93ed50145545b12
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000027_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26cae42918d00675ecd3bd1090c2bd481c1e351fe9548be76726b18d7dd48564
+size 1871812
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000028_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000028_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..4b3170262046f948e4e83201316a2a8786c4c85e
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000028_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:926fdbb71e0e25f0d570e87e57334142a636549274394850a509423139be475b
+size 528556
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000028_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000028_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..f0f780a5688034b5a19b690f7b7841f0a8d7ceb7
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000028_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f31e791cf7e26e5e57757c1e7c5f961a251c25ed0bd5010b53cd27c6fd51993
+size 8854
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000028_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000028_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..c0bd5914294b40206644c3db3147b8ee8ddec800
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000028_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f55d39c5a88d94f8d69153c54e7ae3918959d0eb63cbdeb33936ba943e18ad9
+size 2064753
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000029_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000029_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..f7f997b9b40f5e940484c19576347a9461576ff0
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000029_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf6cbba25b40bae2e7263880f6173e7b712ee0468a6e3f5d2cb36b468c474709
+size 1519201
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000029_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000029_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..1253471f4753415ba3a88e0dd8fa69aafb5b743f
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000029_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4429d88be524d42527a6e6275b2b853a95be0bf6051c508b75a84862e01b1036
+size 10539
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000029_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000029_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..bd11944ef08c18f48ef14527d874d6bfdaa87cbb
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000029_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f3dab797f9260291cb80ee47f016588c6a9c203479210a89dd24527a0bf5b7a
+size 4395447
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000030_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000030_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..51dd923fd5197c0eb47aca577a7edce11936a0bb
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000030_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f4105b32b32c4d1b590754eb8fbe40d67adc11ae5f17e4a743b1cbfd1460c46
+size 905901
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000030_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000030_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..de6c3062c5407f76591e82730555f1caf45ecef6
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000030_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:813ea5a7e41fa5d692117648a6bb0b576cfaac13395e8c25dc82e68581b0f68a
+size 7547
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000030_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000030_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..b27bb7e0c60330a43a44e45815f10572044d68e2
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000030_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9cb91a2d9bb350f6f66fc37cdf9490fe14b79acbb9f3de58388a7fe5ed1a9a7
+size 31744
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000031_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000031_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..9f1bbda28d4603a3bb8a8621f315b739520fb99d
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000031_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6ff0e15e2adf4691a1c2a97ad8deb8c34e3dff41981b9b5a35214bc15093abc8
+size 1160203
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000031_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000031_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..ba977a13181040502ec5f21f7c8eef8df346c653
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000031_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e327b7dc847ea9acf835a8f496323b3de62655793451a92d185779209fe56095
+size 7810
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000031_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000031_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..0775d00d6156856603544245d528f17ffc0320cc
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000031_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b3cceb35c9e86780129b347e120359409d25c9d8fa004b1d8d4efb1d3b07ae0
+size 33914
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000032_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000032_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..1bb37254c8f5e63b076450fc04b4e465acc6c89b
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000032_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ee1db6f6d6660591c447d48573aa30ff94224c66df9694bab628e31e69915a6
+size 1272697
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000032_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000032_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..7b85bd7e4748dbca675c5030195e49fe12e1f96b
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000032_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a4212c75b4d3ed3d3d3e552d166c7efea8fa58ab3982f8ea35934813aa22c1f
+size 12730
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000032_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000032_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..865bf86fe341c5ac4af9160c439b5409ef374a45
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000032_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ebdc64fcdbeeb1949a8cd42cedfd60147968a1955eee4557063b6f3e7b007798
+size 55899
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000033_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000033_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..7abd6695c1a14c755985251ba57b232099971884
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000033_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af7ef4b4d87f296e0528c4a26347a5be186ec11f0b980118e0e4135b4c37a2e3
+size 655455
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000033_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000033_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..255fbb605dc698702ed93e9bf9181b60743f6dfc
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000033_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:674b1a53eb73d3e726cf076c1f272878eff6e2e04df0dc80e01fe4f5b8570899
+size 8246
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000033_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000033_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..65b7ea43b654d9a744276ab22583b91f057dbf8f
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000033_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16ceb1e076fdafd1e375c85e48910015bdf1032385245c0375f2edb847c035f8
+size 769152
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000034_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000034_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..99bf5efcf607cee42e6177f597034f8a79b86a0f
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000034_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b6c082f077c534969bc659f271b722ca22591bb249aea68d9e7c89081f672c0
+size 1020943
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000034_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000034_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..886447de7ef73b1fee91fcddbcb0760313d9d57f
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000034_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:625d60825ccd2fd75f7d9b188a4dd2ffe802d373cc27db141368a998f04ce99a
+size 7355
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000034_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000034_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..69c52f59e2931841e49705db17c86e47d7bfe325
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000034_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5787ef2ac251e2cfe6548ce892ae684ec3a904b794896d0de063ed5730f9ff37
+size 53940
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000035_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000035_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..7744a732f437bc366643e7613d955359fbdde5c1
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000035_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a20d57442c3f804ff3f7e6f0978d278b340ca3fda088eb6581a5c63d8f6de9d
+size 813137
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000035_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000035_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..433e9849466779d86a5685092f1344ea5c100716
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000035_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f0ae2ee0ace188c0d6b8c07b55133a331314cc1cebd812dea657d6f0c77d8c9
+size 8514
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000035_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000035_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..1a2f4c33f8f067ab37dc6b07fe146819c8edcc34
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000035_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e38dec601aa2f3c4afdefe5565b2b32626a5459d7bbdafec6dc1f64de3a0062
+size 911044
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000036_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000036_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..929eb388462e341cbf7308a26c2d3afa4e2a721e
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000036_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b646a4c986044f7c9e82f19b25f66b65229f23546ba56c92c942d921e8403736
+size 611090
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000036_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000036_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..5c3ca5e32a20752c386f3c2b83de56222e834d76
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000036_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b41b521ada82ef1f924abac8e4d666165f02095b0f15ee9e45445a3f82b72b96
+size 7258
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000036_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000036_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..aecf6faa015a0a5a7195eaa81ffbe6bdad01a239
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000036_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:601c34c2d33b8c0e9d735701b97a71f14b50fefa8e46bc991a785ec7cc9967e3
+size 891626
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000037_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000037_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..678cf6da5ec4a109152506831ade1e8599b06abc
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000037_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:80ee61808e63d2fbfcf7d9ead14f5f85130832f282a895b4d12ea45dce65f62c
+size 580252
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000037_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000037_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..e07fbf88aff8a2fe245633b92868b01c13e82045
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000037_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0517aa0e366186adc148f78e6cf2a3154dcf7cd39f96b893ca96fe8788e29b9
+size 6828
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000037_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000037_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..e05e55d70b53035ff16503a05a9facf1df4dfdc3
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000037_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06eae01b924b17d2032d7fc36a1f94734494898565905ed6d6b4cb4abe9b4a0b
+size 955886
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000038_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000038_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..8331d6967eeebee5c58a34bfbf459febd12a2831
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000038_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c537f16fcb34ab40ec318f6ae4874fca347d9d5f18fd61b7422c1df3b8c0a96f
+size 554405
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000038_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000038_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..81358cbd014556da94b124374b1f84e0b4a58ed0
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000038_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:321677c978982af572a7922c9a7dbdc2883c57dd83fe401316dd5d6d1e2abec7
+size 15694
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000038_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000038_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..c16dcc899e2fd12dede0ae27eea27c3b3915443f
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000038_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a777411f830f67e34c3cbac3b6003f4acc7cff888475177bb0745d82f93a914c
+size 40889
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000039_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000039_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..a146660c672fe55ba95966b28783e9bc0dbe5693
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000039_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10910055510f7508d557ff76e1a8004e2d34e42baa76f86d7f4eb934357af8f8
+size 1232747
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000039_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000039_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..c1c5c740a3683562ff882e41654eb862f96a897e
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000039_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a93583de5cb42c1abdf484e622a84140d7a4b8bf179ae4bb4ead8aa8be5ef51
+size 7859
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000039_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000039_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..bb610b3df48f9880710530f90465d0ef425e1af6
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000039_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e514a4c80256f0c486391992ff917b9d87e940bb30a9782119051fd0c8e65ebd
+size 60790
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000040_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000040_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..4ec7c4e3bdf0bdb9492c7aa6a3086feac90b32d3
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000040_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd83617129049ce7f82c57c88f9d66cadd4ba2567393610cca69dbcf12ac9ad9
+size 661517
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000040_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000040_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..ef14ce63875db3f33a983731bf2e935e6d595110
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000040_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6ad4d0eb63c614acb0bb38a508557878a887bec6e57423a0ab8fa1df8b91ace
+size 6655
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000040_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000040_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..545da1e84cf4a0301454a152bf59ded8ef851888
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000040_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d8cb32206fe23fe954711e65183b15f8d70927dfb934d43b0f53d7a1d7265d4
+size 487794
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000041_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000041_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..10d84fa46555b800e805e4d94fc144d740d538d1
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000041_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93d51a7afb157f1135a79789b0047f9c9d6e795231e5ebbeae8346015e8c6ec9
+size 656777
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000041_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000041_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..26686ff6eba9e79d9f9c4c62fb35a0e60b9e24ca
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000041_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:711b28f1cbe1d8c97e7956a858976d82bf3a510da1f0f965b6144782f0f4e55b
+size 7233
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000041_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000041_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..03f41ffa29457e8aa3be5c08f7c9501d7301e06a
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000041_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fff40f17a2f556b3235b937de94ec55ab06b10159cde8afb0131f673ce203762
+size 2852132
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000042_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000042_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..0ae8945c862d45d38c3f76a9e237253707669bc5
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000042_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a01b4a32b5bf574e9c9119ef157ff002a1b3ec34c0fbe76c70bd9c7ae4bfdb4
+size 1195907
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000042_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000042_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..9f76de250c3d5a21e2525f29fbf9436719915a7e
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000042_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e50f99e5ab384b2e77066ff2ee155554fca2ddb80467d5d19bcfb5a58132ca88
+size 5413
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000042_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000042_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..e1e8a6d526c59a017ed820ddf863e7cccdadd1a6
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000042_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:59edbfb4068a085ee27fc0167ac6f65c92a7f577ec861298b1ac7dc902d7766e
+size 1034892
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000043_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000043_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..92fb5198283d170e2387517bd82dc4ff626edc5a
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000043_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ca4c8994ff11a1a0a927e9d1556d6fa3a2185c6572b9167bc457023b3891212
+size 1471505
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000043_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000043_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..35d86a997b86222f7ba4bd35ab011896ce1f1a3f
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000043_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3f408dcb9b8043d4061babaff0f3b270792ac6fe20bf93758e3eeb85566d674
+size 7954
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000043_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000043_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..fd7e99d12d67737e9c74edd8b24787dad521d761
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000043_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:047cb8b822505cf55c8f588b3fc29c214a9309c15176c9d7ae5b48cc2f4086dd
+size 1085721
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000044_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000044_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..2a176ff08b630cc1808d3335946cb716be7d2ef3
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000044_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:783aaf63357672452453cbff6ca8f170b9e035563a1ac53f72d326a3dbd82bc4
+size 3186733
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000044_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000044_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..e4afa545762ee733d20fab505105b88b9fe3d702
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000044_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:533891afe137330c186dc0f838ab308c19273f736832d1c98dec7b3ba0ca8029
+size 8060
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000044_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000044_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..b06c98de72e81b49059dd764ba4671b1975d0dbf
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000044_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3dc069fda4e292200f2fc543391e3cccdeaaf1fe0c092366d425d3d23109cd32
+size 2891726
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000045_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000045_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..4276a28504b6ae1eca343d81f69c761e4bc6126a
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000045_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9ad1fdd2ec0d89310300a01275f4e8100eee82f2bc58d61e4a94caca7d797ae
+size 1626806
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000045_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000045_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..3524078cf12bff17304e08662fbd86b25f9fc4a8
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000045_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bbeea48e0fd3aa92d559e9e320f5fcebc66ed4d0d15a21c5ecc7711f845b466d
+size 5884
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000045_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000045_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..c76f231189b492aebc73776611b9ff55b3f5dcb6
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000045_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f3c8fef2c5e84cf39433bdc994b1fde4c7d7f6cb7a7cf32cec2d8a2d3ca04a3
+size 1273542
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000046_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000046_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..d0058254abd1c5e0a5f927e99f740ab0db8208f9
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000046_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de8484524d1706e1ccdb9a5fd10722f2747ce081ec17db39ca9cb5cd0c877611
+size 1472889
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000046_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000046_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..c9702888167e153d287fa89989123908aafc65f2
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000046_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fafaf8d566333ddc3dc064840bf3082bd98c91837836b58bb6092efeb7e66de2
+size 7481
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000046_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000046_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..603dbb295a0baf9a44caaf8552ca5527d84c8cc5
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000046_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8986364bae7015191b25660981ce07ff54ba575fe12db2cf035c942df3f88880
+size 1209379
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000047_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000047_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..c61d5e21ec5c7ddc2ca9aee5b2aaeac3a1ba5e6f
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000047_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e2fa2a226c0ca98dbdcfca7d5fd405dbf7c902397cdacd924701783501b78a5d
+size 731303
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000047_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000047_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..7e142fa51a00282a6ca451c64a1b176b08e85771
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000047_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57e73a2e6d67af5d5c809e9839c5deffc7e244889af3416e5eceedcfa94857c3
+size 7804
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000047_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000047_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..6d10632d0a2d5f60bf62ec6f81c4497d8372f4c1
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000047_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4388627c6f61feb3e95f0e90dc59ecc0e1119afcaace4bc228e2d84f8b5f2864
+size 585658
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000048_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000048_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..bb0cde72fcbfe714f24b328f6f1e448887a9efa0
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000048_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fdc06e46cd2ebb513ea9fa425249f72a6e88806b0c3dc752e7a9f8f38aacc806
+size 1655392
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000048_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000048_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..02c3211f57378ad1bab41a20cf16dd2e1145703d
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000048_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb4a2b3326730c690878829395c70010c8b1217a971f544cfbc83566efbc4231
+size 9194
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000048_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000048_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..c0536f23713076b657f0f007cada77c627ebed14
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000048_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53cff43d33638596cddc323538fb08d444a1e4bd81a99c11641adcfe3be3bfff
+size 1408575
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000049_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000049_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..ff6b74322d17e7f66f5cd131b13d2c74a785ef3f
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000049_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:39529917053423a8f2ebff35530d2eac1b91a080fa4efded6bac6bf01e9d1d46
+size 1270875
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000049_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000049_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..d9e171aa7a083a176d944a5fdf5b09c7fa845fe4
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000049_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e3f03fa9cf6960b70a267893ae5473b783a0bd55660d18e8b0fcc981e8ae2b0
+size 6977
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000049_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000049_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..a72c0816a15b8cb974ac7fd9dba9f080feab66d2
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000049_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6423395cd8a876844bffa542ebde83b85fed38b26874811f70270592d773b7d
+size 967269
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000050_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000050_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..84e82d58c93f0ce8b453a7c2da12641161851cbb
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000050_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c316c65d3cf8c71d3ef7c96e7da4b533532bd5f3f6bc9e9aeb01eefd69f27fb0
+size 505825
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000050_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000050_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..b5d6e312f4c98be83e005bfc6eb95533cc292124
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000050_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:abd992f82678bbd6035a2bb258d8bbcd253a292127465bd4c5fd0c427175d574
+size 7274
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000050_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000050_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..1664260f83d04da5a6785e8af3e128e68a444024
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000050_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c75e8db8165f742983cd904a85a8fb1f0d0808c75d83e0570f676715c73fa0d6
+size 338474
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000051_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000051_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..406357cd5609ba2c1754ca2af6261793b7b70608
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000051_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04fe7622ad9e85fd8cc2d33c9183255d89ececf623c299f357c4c46e467dbbcd
+size 3280151
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000051_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000051_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..4a72b192835d8e68d83697e7ca7b2f9062100298
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000051_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b6f3d2681aed8d38505bea1c70ce30918134859c6d67b18bb14f50c7c5d8e18
+size 17228
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000051_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000051_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..f70662830ae94853011219fe91302308fae7036b
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000051_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ab7c22cd4bbdfdb6a910186fa4f74e30ec60647eadf63659d98d8464f45292b
+size 967356
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000052_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000052_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..8b4b14d9b55741f23f798ea87a71efacdd35d0c1
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000052_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f369865afeec731a944701f5f3198d2c7508834ca4338689ec158ec4bf3fb734
+size 1870450
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000052_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000052_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..e2747ca8730fb9aedfaecbe48a322f53814b6af7
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000052_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e707dbe835594003502918494f085b2c57ffbb0462eafd400bd75acca603520
+size 8161
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000052_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000052_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..ffbd17eb55df7e78fe8acb0e4fd217139a40540d
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000052_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e68bbd5eb51630be942db0f9177a35a77e7930af252a94cfb66c4f3d68d4edec
+size 827120
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000053_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000053_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..3f7c50ab0b370b4d19353300c44d9180583287cd
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000053_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6eddb8f856a7ec8ba915c4af371cc3545c3b2ea34c8a1030ae41301887b34093
+size 893701
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000053_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000053_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..0e7d2a302f8cfdaa789f528d6e146c79c3b53bf2
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000053_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e9919e9946e2d4c7952a16c1da676c88bdd9724b2b8d81d8c66b0b2a480c4fcc
+size 6149
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000053_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000053_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..bea6805bd42f19526da9c8a9f6c882a3c69086cc
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000053_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dec16429132e75a3ca112def2887f553b3fb06aead92f0fda15145a47a78a75e
+size 313782
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000054_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000054_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..4e0677ada983a345a286e0eaff2bea2f780f953d
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000054_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a61d814f5c6b7eb552aca9d70deaa7b9fc64655531155e69e0352391ee265d1
+size 1402414
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000054_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000054_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..7235be6502a5cfa4b67cfc61329ce2c29312483c
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000054_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7fb0138589dcba7c56692beef8b9d7168da6b7caa11cae3ab2fefc648835fb48
+size 8062
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000054_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000054_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..31739abd751bcacfafc8d70348fff5a20f93b3c0
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000054_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e40bdd65aae49115daf7b29958e776ef3b4aa6017dcaf408b3739d1e5fd99f4
+size 616777
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000055_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000055_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..ebfaae7ba8d0d5a1cda4743215e1ccf272cf2025
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000055_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:799d4143e948a89adba595024d2846d963ee20883974a40d48a435a6c2b271b5
+size 1345175
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000055_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000055_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..2f5abe65bf998705bfd96f17fccd2ea0eb67dc8f
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000055_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3283db493cd944387beacab5de85819a27a346072044eed6f8aacb67ac36e58
+size 7647
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000055_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000055_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..4cae0e2b20af0c3092b4dcf34f039f1308094284
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000055_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e4168625c20bed08bd54830813165bfbf87c8b8b36c9af88192149586cf5a799
+size 548160
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000056_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000056_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..339551e8fde26810f46a2c18ea8bcfaf66fbd0bc
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000056_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:80716cb93a68ee7ecccd9e8e2ff87881d855dee5d828722a9d088e63235a1e98
+size 924302
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000056_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000056_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..69228637daf72120a60dd01e395bcb68e0a9ac6d
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000056_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ab9fb2e0f80c1d700f5f8a88bf1335388771249e24d69c95a1794ee857e6dcd
+size 8837
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000056_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000056_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..ff26aef4ec84183beb6ea220a0a42fbba16fc318
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000056_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a88ddc2b24fa4e3830138f272a5991df91402ac3c995f3c01ee708dc06b55763
+size 494877
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000057_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000057_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..ec8aa9eb6a6b5c5b9fc950d52c5c8cdfc290f73e
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000057_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8c54c7bcb074bc31592ee58e3c271c237405d1dc63f1f1dd7649fe621188bfa2
+size 1695876
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000057_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000057_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..f8cc56b686d97f7ebbbd56ba016990c9cf1bf0c7
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000057_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:870e432224b71c94870e44684e1658d789554d36f7e73d8a3adb4133ef4708f6
+size 7388
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000057_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000057_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..1b2015815a089597acf5f71a0650eb2dea09eeac
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000057_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68a50e019fa4c559082445a85c04a3b51bef1856521e55791f954c0793169363
+size 540713
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000058_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000058_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..6f0c6a22b16a71d197339810cd65c48b16e5f0c4
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000058_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f8e9bb1083beccce88ba57f48e632e29460e496d4345c6bd9a998cd9d9f9878
+size 895498
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000058_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000058_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..2aa633124f9330dafa447e28b0ab0693ceaaadbe
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000058_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64db79291ce2c7c807a8e2c20a842e187521ac3905e15163acea489e962aac1b
+size 8229
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000058_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000058_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..5de224bfdc7ca8f9aa00b61bcc295ee54eb57b37
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000058_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d20f6416cab0a1f3bacc6c7a4dbf64cd1b3441427b74200e9c1ea32f41d5ce1c
+size 2757945
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000059_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000059_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..2e283b35a90e63d47b82cfa816a0310b133c942f
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000059_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e916bda852ede55f2672a7504bab677c1c96d810aa5cbdbe162c15e7f7de0bbb
+size 677633
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000059_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000059_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..b5d6e312f4c98be83e005bfc6eb95533cc292124
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000059_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:abd992f82678bbd6035a2bb258d8bbcd253a292127465bd4c5fd0c427175d574
+size 7274
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000059_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000059_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..68d2f482ae3c69c3199b0e236363f0677c8494ff
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000059_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:102e76718bccfc20f7d26d1158f55a022e8a108b0aa32ce6599edc599df6d9a7
+size 400702
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000060_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000060_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..1842f2041f3f27f2cc6f451954794837fbed7c65
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000060_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a736ef440920959124dbe80ab8ea37d0bb877af0a4fa6aabed4ac4fdc9aa356d
+size 1162103
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000060_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000060_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..8c50ea48072245c736f15d6f8027218a9a7ae218
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000060_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dfa30aa09c9df554ad98a9af1901f34780d74b9ad6c7aa6bcba781bb3c99bbea
+size 8287
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000060_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000060_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..76be6af2a61d71851939d0f2513a15f3e29afc0f
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000060_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:42ec375693b9a0132626b30f5385465e114d80f44d452e4c624d3445b1c504ad
+size 18751295
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000061_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000061_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..9bcc0085616cddebd945e215352c7ed28c83ccfa
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000061_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee7c91adea9cfb3168d38814eeb11a7d0f4014d060791da52e516ce3ac244bf6
+size 701912
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000061_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000061_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..aaf169cd00f82144b7cf046b1d947ff835a900b6
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000061_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa52c842e33bbad18829b7b41d85ade22975684a126b5f9fd6f5195d04535e84
+size 7181
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000061_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000061_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..e4fc382987a86328a8756dad415d234fdc48b78e
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000061_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:978eb9177e9e82357b579ce6496f6c347776cf9955ef6a4e2a5490329069c3b1
+size 2810115
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000062_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000062_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..5f88864a8c4c494e006a085be5eea35babd698b7
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000062_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f058f3d386d482afec926891fdfd11161d3a6326413a83d8fa9b936aafc5a6d
+size 710541
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000062_src_mask.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000062_src_mask.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..f88ed8b27076edcc4132e09e8106ecc8a80cb609
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000062_src_mask.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09fe6eac55e3fcbc6ad3e806e91fe022cdce8efd6bebd7411fc3a4888220523b
+size 6818
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000062_src_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000062_src_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..a28cf4e9118f0022e9947d605977462c503a2935
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000062_src_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a89ede00fedb93a66657835538de4e8c44224344c9148957856444cfa94331c
+size 3156427
diff --git a/video_gen_14d/data/VACE-Benchmark/data/sample000063_raw_video.mp4 b/video_gen_14d/data/VACE-Benchmark/data/sample000063_raw_video.mp4
new file mode 100644
index 0000000000000000000000000000000000000000..753ae432fccb2430ffb0dda6d2e5c45ac2e4b1d3
--- /dev/null
+++ b/video_gen_14d/data/VACE-Benchmark/data/sample000063_raw_video.mp4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:54330cdc25bf24abcffdbde8358a3bdb4789375a02b26531463fd9772f256535
+size 1308641