wan2.1

Wan2.1 is an open and advanced large-scale video generative model suite. It supports multiple tasks including Text-to-Video (T2V), Image-to-Video (I2V), First-Last-Frame-to-Video (FLF2V), Video Editing, Text-to-Image, and Video-to-Audio. The repository provides model checkpoints for 1.3B and 14B parameter models, supporting 480P and 720P resolutions. Key capabilities include: - Consumer-grade GPU support: The T2V-1.3B model requires 8.19 GB VRAM, enabling video generation on standard GPUs. - Visual text generation: Capable of generating both Chinese and English text within videos. - Wan-VAE: A video VAE for encoding and decoding 1080P videos of any length while preserving temporal information. - VACE: An all-in-one model for video creation and editing. The repository offers single-GPU and multi-GPU inference scripts using FSDP and xDiT USP (Ulysses and Ring strategies). It supports prompt extension via Dashscope API or local Qwen models. Integration with Diffusers, ComfyUI, and Gradio demos is provided. Model weights are available on Hugging Face and ModelScope.