datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
wan22-processed-clipswan22-animate-3k-opensource-data
Wan2.2 Animate Open Dataset Pack
This dataset repo stores the complete datasets/ directory used for the Wan2.2 TI2V 5B + One-to-All animate experiment.
The original tree contains more than 10,000 files in one directory, which Hugging Face git repositories reject as raw files. Therefore the dataset is stored as split tar shards.
Restore:
cat datasets.tar.part-* | tar -xf -
sha256sum -c SHA256SUMS
After extraction, the restored tree contains:… See the full description on the dataset page: https://huggingface.co/datasets/simbahuang/wan22-animate-3k-opensource-data.MiraData_Wan22_Latentswan22-physics-videos
Wan2.2 Physics Video Generation Dataset
A dataset of AI-generated physics simulation videos with intermediate denoising latent tensors, created using the Wan2.2 T2V-A14B text-to-video model.
Dataset Summary
480 videos (.mp4) generated from 60 unique physics prompts (8 random seeds each)
Intermediate latent tensors (.pt) saved at every 2 denoising steps (10 snapshots per video)
Physics categories: ball bouncing, pendulum motion, objects sliding on inclined planes… See the full description on the dataset page: https://huggingface.co/datasets/hivamoh/wan22-physics-videos.wan22_datasetcomfyui-wan22-assetsvbench-wan22-t2v-dense93-xpu
VBench dense93 — Wan2.2 T2V videos (Intel XPU / vLLM-Omni)
93 text-to-video generations for the VBench "dense93" prompt set, produced
with Wan2.2-T2V-A14B (Diffusers) served by vLLM-Omni on Intel XPU
(oneAPI, 720x1280 @ 16fps, 81 frames, 40 denoise steps, guidance 4.0/3.0,
boundary 0.875, flow shift 5.0, seed 42, MXFP8 linear + cache-dit +
Sage V3 hybrid attention with SDPA fallback on blocks 33,34,38,39).
One video per prompt; file name = <prompt>-0.mp4
(prompt list:… See the full description on the dataset page: https://huggingface.co/datasets/Yi30/vbench-wan22-t2v-dense93-xpu.wan22modelAmputation_Wan22-datasetwan22_datalorawan22-rollout-put-object-cabinet-step0
Wan2.2 TI2V Step-0 Rollout — RoboTwin put_object_cabinet
160 video rollouts (10 scenes × 16 samples) generated by Wan2.2-TI2V-5B + merged Vidar LoRA
on the RoboTwin put_object_cabinet task. These are the pre-NFT-training (step-0) baseline
samples used to evaluate reward-model behaviour and seed RL fine-tuning.
Generation config
Base model : Wan2.2-TI2V-5B
LoRA : vidar/merged_vidar_lora.pt (vidar baseline merged into DiT)
Sampler : deterministic ODE (Euler… See the full description on the dataset page: https://huggingface.co/datasets/VincentNi/wan22-rollout-put-object-cabinet-step0.wan22-sampleMultiCamVideo-Wan22
MultiCamVideo-Wan22
WanPRoPE preprocessing of
KlingTeam/MultiCamVideo-Dataset
for the Wan2.2 TI2V 5B backbone.
Contents
136,000 rows from 13,600 synchronized scenes and 10 cameras per scene
32 logical shards, with scene groups kept intact
2,144 Parquet files: 67 files per shard
704×1280 center-cropped conditioning, 81 source frames at 15 fps
21 latent/action timesteps
camera viewmats with shape [21, 4, 4]
camera intrinsics Ks with shape [21, 3, 3]… See the full description on the dataset page: https://huggingface.co/datasets/H1yori233/MultiCamVideo-Wan22.wan22-comfyui-models-backupwan22-code-backupwan22-rollout-put-bottles-dustbin-step0
Wan2.2 TI2V Step-0 Rollout — RoboTwin put_bottles_dustbin
160 video rollouts (10 scenes × 16 samples) generated by Wan2.2-TI2V-5B + merged Vidar LoRA
on the RoboTwin put_bottles_dustbin task (put 3 bottles into a dustbin sequentially). These
are the pre-NFT-training (step-0) baseline samples used to evaluate reward-model behaviour
and seed RL fine-tuning.
Generation config
Base model : Wan2.2-TI2V-5B
LoRA : vidar/merged_vidar_lora.pt (vidar baseline merged into… See the full description on the dataset page: https://huggingface.co/datasets/VincentNi/wan22-rollout-put-bottles-dustbin-step0.wan22-rollout-blocks-ranking-rgb-step0
Wan2.2 TI2V Step-0 Rollout — RoboTwin blocks_ranking_rgb
160 video rollouts (10 scenes × 16 samples) generated by Wan2.2-TI2V-5B + merged Vidar LoRA
on the RoboTwin blocks_ranking_rgb task (arrange R/G/B blocks left-to-right). These are the
pre-NFT-training (step-0) baseline samples used to evaluate reward-model behaviour and seed
RL fine-tuning.
Generation config
Base model : Wan2.2-TI2V-5B
LoRA : vidar/merged_vidar_lora.pt (vidar baseline merged into DiT)
Sampler… See the full description on the dataset page: https://huggingface.co/datasets/VincentNi/wan22-rollout-blocks-ranking-rgb-step0.wan22-ti2v-5bslashered_Wan22-datasetWan22lorawan22-incontext-control-data
Wan2.2 In-Context Control — derived training data
Derived metadata/pose NPZs for the in-context camera + audio control fork of DiffSynth-Studio
(training Wan2.2-TI2V-5B). This repo holds only the small derived files needed to reproduce the
camera and audio (e11h) runs. It does not rehost source videos — those come from the original
datasets linked below.
Companion code: the DiffSynth-Studio fork (see its README for the full reproduction walkthrough).
Files… See the full description on the dataset page: https://huggingface.co/datasets/Haosonnn/wan22-incontext-control-data.wan22_i2v_naranwan2_2_i2v_model_ClothConsistency_Videoswan22-vbench-mini10
Wan2.2 dense93 mini10 v1
用于快速比较不同推理配方的 imaging_quality；从已有 93 条中挑选 10 条。
这是自定义筛选集，未经独立 datatype 或 seed 的泛化验证，不是官方 VBench mini benchmark。
文件
prompts.txt：10 条原始英文 prompt，每行一条，保持原 dense93 顺序。
prompts.json：VBench custom_input 的 {video_filename: prompt} 映射，seed 对应视频索引仍为 -0。
selection.csv：分层顺序、原始编号、公开分数、挑选理由。
dense93_scores.csv：从 issue 评论提取的完整 93 条公开分数，用于复核。
analysis.json：分层规则、统计、来源、原始版本和 prompt 校验值。
VBENCH_LICENSE：VBench 的 Apache 2.0 许可证。Prompt 文本未修改，列表经过筛选。… See the full description on the dataset page: https://huggingface.co/datasets/changwangss/wan22-vbench-mini10.Wan_2_2_Xiang_Gen_Images
https://huggingface.co/svjack/Xiang_wan_2_2_lora (lora weight 1.0) + https://huggingface.co/svjack/Watanuki_Kimihiro_wan_2_2_14B_lora (lora weight 1.5)
wan22-penalty-lora-datasetwan22-calibration-promptsWorldArena2-Track1-Wan22-Baselinewan_2_2_s2v_Mavuika_Ad_videos
wow1-robomind-longtext-1400k-wan22