I2V-A14B MoE design
According to the model card, the A14B series uses a two-expert MoE design with about 14B parameters per expert and 14B active parameters per step, and this repository holds the image-to-video variant.
Open Source Model Profile · Wan-AI
Wan2.2-I2V-A14B-Diffusers
Wan2.2-I2V-A14B-Diffusers is a 14.29B-parameter image-to-video model from Wan-AI. According to the model card, this Diffusers-format I2V-A14B MoE model supports 480P and 720P generation and is licensed under Apache-2.0.
- Publisher
- Wan-AI
- Task
- image-to-video
- Model type
- Unknown
- License
- apache-2.0
- Library
- diffusers
- Publication status
- Accepted · not indexed
Model overview
Wan2.2-I2V-A14B-Diffusers is published by Wan-AI as an image-to-video model in the Wan2.2 family. Safetensors metadata reports 14,288,901,184 parameters, and the record lists the diffusers library with an Apache-2.0 license. According to the model card, this repository holds the I2V-A14B model for image-to-video generation at 480P and 720P, built with a Mixture-of-Experts architecture and integrated into Diffusers.
Recorded capabilities
480P and 720P support
According to the model card, the I2V-A14B model supports both 480P and 720P resolutions with more stable synthesis and support for diverse stylized scenes.
Diffusers integration
According to the model card, Wan2.2 T2V, I2V, and TI2V were integrated into Diffusers, with this model loadable as WanImageToVideoPipeline in bfloat16 on CUDA.
Documented local inference
According to the model card, single-GPU inference at 1280x720 can run on a GPU with at least 80GB VRAM, with multi-GPU FSDP and DeepSpeed Ulysses commands documented.
Use cases in the source record
- Image-to-video generation at 480P and 720P using an input image with optional text guidance, as documented in the model card's Diffusers workflow.
- Prompt-free video synthesis from an input image alone, which the model card describes as supported with optional prompt extension.
Limitations and unknowns
- No evaluation results were extracted from this record beyond publisher-described validation-loss comparisons.
- The model card notes this model requires features available only in the main branch of diffusers, not the latest stable PyPI release.
- Provider state is historical snapshot data, not independently refreshed current availability.
Source and provenance
Source: Wan-AI/Wan2.2-I2V-A14B-Diffusers
Captured: Unknown. Processed: 2026-09-07T19:34:38.543247+00:00.
Wan2.2 💜 Wan ｜ 🖥️ GitHub | 🤗 Hugging Face | 🤖 ModelScope | 📑 Technical Report | 📑 Blog | 💬 WeChat Group | 📖 Discord Wan: Open and Advanced Large-Scale Video Generative Models We are excited to introduce Wan2.2 , a major upgrade to our foundational video models. With Wan2.2 , we have focused on incorporating the following innovations: 👍 Effective MoE Architecture : Wan2.2 introduces a Mixture-of-Experts (MoE) architecture into video diffusion models. By separating the denoising process cross timesteps with specialized powerful expert models, this enlarges the overall model capacity while maintaining the same computational co…
F001F002F003F004F005F006F007F008F013F014F018F019F020F022F023F024F025F026F029F033