amdenterpriseai/aim-instinct-model-wan-ai-wan2-2-t2v-a14b-diffusers repository overview

⁠Wan 2.2 T2V A14B (Diffusers)

Wan2.2 T2V-A14B is a 14-billion-parameter Mixture-of-Experts text-to-video diffusion model from Wan-AI. It generates 5-second video clips at 480P and 720P resolutions from text prompts, using a two-stage MoE design (high-noise and low-noise experts) that enlarges model capacity while keeping per-step compute roughly constant. The A14B variant was trained on significantly more data than Wan2.1 (65.6% more images and 83.2% more videos). This is the Diffusers-format checkpoint, intended for serving via vLLM-Omni's diffusion pipeline.