| --- |
| library_name: s2pd |
| license: mit |
| pipeline_tag: image-to-video |
| tags: |
| - video-generation |
| - diffusion |
| - safetensors |
| - evaluation |
| --- |
| |
| # Serial-to-Parallel Diffusion |
|
|
| [Project Page](https://jefequien.github.io/S2PD/) · [Code](https://github.com/jefequien/S2PD) |
|
|
| ## DiT-B |
|
|
| Pixel-space models trained from scratch. |
|
|
| | Dataset | Bidirectional | cDF | BC | S2PD | |
| | :--- | :---: | :---: | :---: | :---: | |
| | Conway | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BD/dit-b-pixel/conway) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/CDF/dit-b-pixel/conway) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BC/dit-b-pixel/conway) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/S2PD/dit-b-pixel/conway) | |
| | Chess | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BD/dit-b-pixel/chess) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/CDF/dit-b-pixel/chess) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BC/dit-b-pixel/chess) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/S2PD/dit-b-pixel/chess) | |
| | 2048 | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BD/dit-b-pixel/2048) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/CDF/dit-b-pixel/2048) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BC/dit-b-pixel/2048) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/S2PD/dit-b-pixel/2048) | |
| | Fifteen Puzzle | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BD/dit-b-pixel/fifteen_puzzle) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/CDF/dit-b-pixel/fifteen_puzzle) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BC/dit-b-pixel/fifteen_puzzle) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/S2PD/dit-b-pixel/fifteen_puzzle) | |
| | Tetris | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BD/dit-b-pixel/tetris) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/CDF/dit-b-pixel/tetris) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BC/dit-b-pixel/tetris) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/S2PD/dit-b-pixel/tetris) | |
| | Snake | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BD/dit-b-pixel/snake) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/CDF/dit-b-pixel/snake) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BC/dit-b-pixel/snake) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/S2PD/dit-b-pixel/snake) | |
| | Rubik’s Cube 3D | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BD/dit-b-pixel/rubiks_3d) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/CDF/dit-b-pixel/rubiks_3d) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BC/dit-b-pixel/rubiks_3d) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/S2PD/dit-b-pixel/rubiks_3d) | |
| | Double Pendulum | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BD/dit-b-pixel/double_pendulum) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/CDF/dit-b-pixel/double_pendulum) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BC/dit-b-pixel/double_pendulum) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/S2PD/dit-b-pixel/double_pendulum) | |
| | Three Body | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BD/dit-b-pixel/three_body) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/CDF/dit-b-pixel/three_body) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BC/dit-b-pixel/three_body) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/S2PD/dit-b-pixel/three_body) | |
| | Colliding Balls 2D | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BD/dit-b-pixel/colliding_balls) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/CDF/dit-b-pixel/colliding_balls) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BC/dit-b-pixel/colliding_balls) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/S2PD/dit-b-pixel/colliding_balls) | |
| | Colliding Balls 3D | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BD/dit-b-pixel/colliding_balls_3d) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/CDF/dit-b-pixel/colliding_balls_3d) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BC/dit-b-pixel/colliding_balls_3d) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/S2PD/dit-b-pixel/colliding_balls_3d) | |
| | Pong | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BD/dit-b-pixel/pong) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/CDF/dit-b-pixel/pong) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BC/dit-b-pixel/pong) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/S2PD/dit-b-pixel/pong) | |
|
|
| ## Wan-5B-LoRA |
|
|
| Adapters requiring the [Wan-5B backbone](https://huggingface.co/Wan-AI/Wan2.2-TI2V-5B-Diffusers). |
|
|
| | Dataset | Bidirectional | cDF | BC | S2PD | |
| | :--- | :---: | :---: | :---: | :---: | |
| | Kubric MOVi-A | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BD/wan-5b-lora/kubric_movi_a) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/CDF/wan-5b-lora/kubric_movi_a) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BC/wan-5b-lora/kubric_movi_a) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/S2PD/wan-5b-lora/kubric_movi_a) | |
| | Kubric MOVi-C | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BD/wan-5b-lora/kubric_movi_c) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/CDF/wan-5b-lora/kubric_movi_c) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BC/wan-5b-lora/kubric_movi_c) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/S2PD/wan-5b-lora/kubric_movi_c) | |
| | MPMWorlds | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BD/wan-5b-lora/mpm_worlds) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/CDF/wan-5b-lora/mpm_worlds) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BC/wan-5b-lora/mpm_worlds) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/S2PD/wan-5b-lora/mpm_worlds) | |
| | Rubik’s Cube Real | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BD/wan-5b-lora/rubiks_real) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/CDF/wan-5b-lora/rubiks_real) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/BC/wan-5b-lora/rubiks_real) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/S2PD/wan-5b-lora/rubiks_real) | |
|
|
| <details> |
| <summary>Block size ablations (DiT-B)</summary> |
|
|
| Pin-8 and Pin-64 train with their named block size and full-video blocks, weighted 80/20. |
|
|
| | Dataset | Pin-8 | Pin-64 | |
| | :--- | :---: | :---: | |
| | Conway | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/conway) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/block-size/pin-64/dit-b-pixel/conway) | |
| | Chess | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/chess) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/block-size/pin-64/dit-b-pixel/chess) | |
| | 2048 | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/2048) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/block-size/pin-64/dit-b-pixel/2048) | |
| | Fifteen Puzzle | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/fifteen_puzzle) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/block-size/pin-64/dit-b-pixel/fifteen_puzzle) | |
| | Tetris | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/tetris) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/block-size/pin-64/dit-b-pixel/tetris) | |
| | Snake | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/snake) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/block-size/pin-64/dit-b-pixel/snake) | |
| | Rubik’s Cube 3D | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/rubiks_3d) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/block-size/pin-64/dit-b-pixel/rubiks_3d) | |
| | Double Pendulum | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/double_pendulum) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/block-size/pin-64/dit-b-pixel/double_pendulum) | |
| | Three Body | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/three_body) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/block-size/pin-64/dit-b-pixel/three_body) | |
| | Colliding Balls 2D | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/colliding_balls) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/block-size/pin-64/dit-b-pixel/colliding_balls) | |
| | Colliding Balls 3D | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/colliding_balls_3d) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/block-size/pin-64/dit-b-pixel/colliding_balls_3d) | |
| | Pong | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/pong) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/block-size/pin-64/dit-b-pixel/pong) | |
|
|
| </details> |
|
|
| <details> |
| <summary>Token order ablations (DiT-B, Pin-8)</summary> |
|
|
| | Dataset | Raster | Zigzag | Hilbert | |
| | :--- | :---: | :---: | :---: | |
| | Conway | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/conway) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/zigzag/dit-b-pixel/conway) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/hilbert/dit-b-pixel/conway) | |
| | Chess | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/chess) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/zigzag/dit-b-pixel/chess) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/hilbert/dit-b-pixel/chess) | |
| | 2048 | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/2048) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/zigzag/dit-b-pixel/2048) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/hilbert/dit-b-pixel/2048) | |
| | Fifteen Puzzle | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/fifteen_puzzle) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/zigzag/dit-b-pixel/fifteen_puzzle) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/hilbert/dit-b-pixel/fifteen_puzzle) | |
| | Tetris | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/tetris) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/zigzag/dit-b-pixel/tetris) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/hilbert/dit-b-pixel/tetris) | |
| | Snake | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/snake) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/zigzag/dit-b-pixel/snake) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/hilbert/dit-b-pixel/snake) | |
| | Rubik’s Cube 3D | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/rubiks_3d) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/zigzag/dit-b-pixel/rubiks_3d) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/hilbert/dit-b-pixel/rubiks_3d) | |
| | Double Pendulum | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/double_pendulum) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/zigzag/dit-b-pixel/double_pendulum) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/hilbert/dit-b-pixel/double_pendulum) | |
| | Three Body | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/three_body) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/zigzag/dit-b-pixel/three_body) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/hilbert/dit-b-pixel/three_body) | |
| | Colliding Balls 2D | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/colliding_balls) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/zigzag/dit-b-pixel/colliding_balls) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/hilbert/dit-b-pixel/colliding_balls) | |
| | Colliding Balls 3D | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/colliding_balls_3d) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/zigzag/dit-b-pixel/colliding_balls_3d) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/hilbert/dit-b-pixel/colliding_balls_3d) | |
| | Pong | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/raster/dit-b-pixel/pong) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/zigzag/dit-b-pixel/pong) | [ckpt](https://huggingface.co/jefequien/S2PD/tree/main/checkpoints/ablations/token-order/hilbert/dit-b-pixel/pong) | |
|
|
| </details> |
|
|
| Download and usage instructions are in the [code repository](https://github.com/jefequien/S2PD). |
|
|
| ## Datasets |
|
|
| - [Kubric MOVi-A](https://huggingface.co/jefequien/S2PD/tree/main/data/kubric_movi_a): 9,703 training and 250 validation clips. |
| - [Kubric MOVi-C](https://huggingface.co/jefequien/S2PD/tree/main/data/kubric_movi_c): 9,737 training and 250 validation clips. |
|
|
| - [Rubik’s Cube Real](https://huggingface.co/jefequien/S2PD/tree/main/data/rubiks_real): 5,000 training and 512 validation clips. |
| - [Chess](https://huggingface.co/jefequien/S2PD/tree/main/data/chess): 1 million training and 10,000 validation games. |
|
|
| ## Licenses |
|
|
| Original weights, adapters, Rubik’s Cube Real footage, and training curves: [MIT](./LICENSE). |
| [Lichess games](https://database.lichess.org/): [CC0](./LICENSE-CHESS-CC0-1.0.txt). The [mirrored Wan backbone, VAE, UMT5 encoder, and tokenizer](./pretrained/wan-2.2-ti2v-5b): [Apache 2.0](./LICENSE-WAN-APACHE-2.0.txt). |
|
|
| S2PD Kubric renders: [CC BY-SA 4.0](./data/kubric_movi_a/LICENSE.txt); see the dataset cards for upstream attribution. |
|
|
| ## Citation |
|
|
| If you use S2PD in your research, please cite: |
|
|
| ```bibtex |
| @misc{hu2026s2pd, |
| title = {{S2PD}: Serial-to-Parallel Diffusion for Physically and Logically Consistent Video Generation}, |
| author = {Hu, Jeffrey and Olmeda Reino, Daniel and Tewari, Ayush}, |
| year = {2026}, |
| eprint = {2610.06847}, |
| archivePrefix = {arXiv}, |
| primaryClass = {cs.CV}, |
| url = {https://arxiv.org/abs/2610.06847} |
| } |
| ``` |
