Wan 2.2
by Alibaba (Tongyi Lab) · Early 2026
MoE architecture with 27B total params but only 14B active. Trained on 65% more images and 83% more video than 2.1. Outperforms leading closed-source models on Wan-Bench 2.0.
Specifications
Max Resolution
720p
Max Duration
10-15s
FPS
24
Native Audio
No
ComfyUI Support
Yes
Fine-tunable
Yes
Min VRAM
8GB (small) / 24GB (full)
Cost / Second
Self-host
Architecture
DiT + MoE (2-expert: high-noise + low-noise)
Parameters
27B total (14B active per step, 2x14B experts)
Inputs
T2V (A14B), I2V (A14B), TI2V (5B), S2V (14B)
License
Apache 2.0
Strengths & Trade-offs
Strengths
- First MoE in video diffusion
- 27B total but only 14B active per step
- high-noise expert for layout + low-noise for detail
- +65.6% more images and +83.2% more video training data vs 2.1
- cinematic aesthetic control (lighting, composition, contrast, color tone)
- TI2V-5B runs on consumer 4090
Weaknesses
- 720p cap
- MoE needs careful threshold tuning (SNR-based)
- no native audio in base model (S2V is separate)
- newer ecosystem than 2.1
Best For
- Self-hosted production
- cinematic style control
- speech-to-video
- consumer GPU deployment (TI2V-5B)
- academic research
Scores
Workflows on Floyo
Wan 2.2 Animate Preprocess (Kijai)
Preprocessing workflow by Kijai for Wan 2.2 animation pipelines. Prepares input data for downstream Wan 2.2 video generation workflows.
Open Workflow →Wan 2.2 + Qwen V2V Restyle
Video-to-video restyling combining Wan 2.2 with Qwen. Transform existing video footage into a new visual style while preserving the original motion and composition.
Open Workflow →Wan 2.2 T2V with UnifiedRew
Text-to-video workflow using Wan 2.2 with UnifiedRew integration for enhanced generation quality and consistency.
Open Workflow →Wan 2.2 Animate Character Replacement
Replace characters in animated sequences using Wan 2.2. Swap subjects while maintaining the original motion, timing, and scene composition.
Open Workflow →Compare with other models