Type: huggingface.image_to_video.Wan_I2V

Namespace: huggingface.image_to_video

Description

Generates a video from an input image using Wan image-to-video pipelines. video, generation, AI, image-to-video, diffusion, Wan

Use cases:
- Turn a single image into a dynamic clip with prompt guidance
- Choose between Wan 2.2 A14B, Wan 2.1 14B 480P, and Wan 2.1 14B 720P models

Properties

Property Type Description Default
input_image image The input image to generate the video from. {'type': 'image', 'uri': '', 'asset_id': None, 'data': None}
prompt str A text prompt describing the desired video. An astronaut walking on the moon, cinematic lighting, high detail
model_variant Enum['Wan-AI/Wan2.2-I2V-A14B-Diffusers', 'Wan-AI/Wan2.1-I2V-14B-480P-Diffusers', 'Wan-AI/Wan2.1-I2V-14B-720P-Diffusers'] Select the Wan I2V model to use. Wan-AI/Wan2.2-I2V-A14B-Diffusers
negative_prompt str A text prompt describing what to avoid in the video. ``
num_frames int The number of frames in the video. 81
guidance_scale float The scale for classifier-free guidance. 5.0
num_inference_steps int The number of denoising steps. 50
height int The height of the generated video in pixels. 480
width int The width of the generated video in pixels. 832
fps int Frames per second for the output video. 16
seed int Seed for the random number generator. Use -1 for a random seed. -1
max_sequence_length int Maximum sequence length in encoded prompt. 512
enable_cpu_offload bool Enable CPU offload to reduce VRAM usage. True
enable_vae_slicing bool Enable VAE slicing to reduce VRAM usage. True
enable_vae_tiling bool Enable VAE tiling to reduce VRAM usage for large videos. False

Outputs

Output Type Description
output video  

Metadata

Browse other nodes in the huggingface.image_to_video namespace.