使用 Wan AI 2.1 创作精彩视频

Wan AI 生成队列正在处理请求。等待期间,请浏览下方的 Wan AI 展示画廊和高级模型功能。

Wan AI 精选示例

Wan AI Text Generation Demo 1

Prompt: A stunning young woman emerges from crystal-clear turquoise water in slow motion. Her long dark hair flows gracefully as water cascades down her face. Shot from a slightly low angle, capturing the moment her face breaks through the water's surface. Soft natural sunlight filters through the water, creating ethereal light rays and sparkles on the water droplets. Her expression is serene and peaceful, with water droplets glistening on her skin. The background shows a blurred underwater scene with gentle ripples. Cinematic color grading with rich aqua tones, professional lighting, shot in 4K quality with high dynamic range. The movement is smooth and elegant, emphasizing the transition between underwater and surface.

Wan AI Text Generation Demo 2

Prompt: A wide cinematic shot from the audience's perspective capturing a vibrant hip-hop crew dominating the stage. Comprising five dancers with urban streetwear, confident expressions, and synchronized movements, they perform under dynamic stage lighting with side-angled beams

Wan AI Text Generation Demo3

Prompt: In the Baroque style European palace, sparkling crystal chandeliers cast a soft glow, illuminating a pair of dancing lovers in the center. The man is dressed in a black tailcoat, paired with a snow-white shirt and bow tie, showcasing a gentlemanly demeanor; The lady wore a floor length long skirt, with delicate lace embellishments at the hem, creating a light and elegant look. They embraced tightly, their arms elegantly intertwined, spinning and jumping with the rhythm of the waltz, each step interpreting romance and passion. The audience around cast envious glances, and the air was filled with the atmosphere of nobility and the sweetness of love. Mid shot, using stable follow-up shooting to capture every moment of rotation.

Wan AI 2.1 采用尖端架构:

Wan-VAE:先进的 3D 因果 VAE

专为视频生成设计的全新 3D 因果 VAE 架构

  • Unlimited-length 1080P video processing
  • Temporal information preservation
  • Efficient spatio-temporal compression
  • Reduced memory footprint
  • Superior performance efficiency

视频扩散 DiT

采用扩散变换器的流匹配框架

Wan 1.3B

1536D

12 heads

30 layers

Wan 14B

5120D

40 heads

40 layers

使用 Wan AI 2.1 创作精彩视频

Wan AI 2.1 是一个先进的开源视频生成平台,可将您的照片和文字转换成令人惊叹的视频。基于先进的 AI 技术,它能在消费级 GPU 上提供卓越性能,生成专业品质的视频。

用户友好

仅需 8.19GB 显存即可在消费级 GPU 上生成高质量视频。在 RTX 4090 上约 4 分钟内生成 5 秒钟的 480P 视频。

SOTA 性能

体验业界领先的视频质量,在多个基准测试中超越现有开源模型和商业解决方案。

多任务支持

一个统一模型支持文本生成视频、图像生成视频、视频编辑、文本生成图像和视频生成音频等任务。

视觉文本生成

首个能够生成中英文文本的视频模型,具备强大的实用文本生成能力。

Wan AI 高级功能

强大的视频 VAE

Wan-VAE 可以编码和解码无限长度的 1080P 视频,同时保留时序信息,完美适配视频生成。

分辨率选项

支持 480P 和 720P 视频生成,针对不同 GPU 配置优化性能。

开源

完全开源实现,配备全面文档,支持社区开发和创新。

Wan AI 技术规格

Wan 2.1 采用尖端 AI 架构和优化:

  • 采用扩散变换器的先进流匹配框架
  • 全新的 3D 因果 VAE 架构实现卓越视频压缩
  • 针对单 GPU 和多 GPU 推理优化
  • 支持 T2V-14B、T2V-1.3B 和 I2V-14B 模型