---
license: apache-2.0
base_model: Wan-AI/Wan2.2-TI2V-5B
library_name: mlx
pipeline_tag: text-to-video
tags:
- mlx
- apple-silicon
- wan2.2
- text-to-video
- image-to-video
- video-generation
---

# Wan2.2-TI2V-5B — MLX (bf16)

Native **MLX** (Apple Silicon) conversion of [Wan-AI/Wan2.2-TI2V-5B](https://huggingface.co/Wan-AI/Wan2.2-TI2V-5B), packaged as a turnkey, self-contained snapshot for the [SceneWorks](https://huggingface.co/SceneWorks) app. Downloading this repo replaces the previous "download the source checkpoint and convert on-device" first-run step.

## Contents (self-contained, bf16)
| file | what |
|---|---|
| `model.safetensors` | Wan2.2 TI2V-5B DiT transformer |
| `t5_encoder.safetensors` | UMT5-XXL text encoder |
| `vae.safetensors` | Wan z48 VAE |
| `tokenizer.json` | UMT5 tokenizer |
| `config.json` | architecture config |

Quantization (Q4/Q8) is applied **at load** by the engine — these weights are full bf16, so one artifact serves every quant tier.

## Provenance
- **Source:** `Wan-AI/Wan2.2-TI2V-5B` (Apache-2.0).
- **Converted with:** the SceneWorks native Rust MLX converter (`mlx-gen-wan`, converter id `wan_ti2v_5b`), dtype `bfloat16`.
- No re-quantization; lean snapshot (excludes diffusers-format directories the MLX engine never loads).

## License
**Apache-2.0**, inherited from the upstream model. This repository redistributes a converted copy of the upstream Apache-2.0 weights, with attribution, as permitted by that license. See the [source model card](https://huggingface.co/Wan-AI/Wan2.2-TI2V-5B).