Image-to-Video
Diffusers
Safetensors
English
Chinese
vace
video generation
video-to-video editing
refernce-to-video
wan
wan2.1
video-to-video
video-generation
bf16
Instructions to use wavespeed/Wan2.1-VACE-14B-bf16 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use wavespeed/Wan2.1-VACE-14B-bf16 with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline from diffusers.utils import load_image, export_to_video # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("wavespeed/Wan2.1-VACE-14B-bf16", dtype=torch.bfloat16, device_map="cuda") pipe.to("cuda") prompt = "A man with short gray hair plays a red electric guitar." image = load_image( "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/guitar-man.png" ) output = pipe(image=image, prompt=prompt).frames[0] export_to_video(output, "output.mp4") - Notebooks
- Google Colab
- Kaggle
Add model card metadata: base_model, license, pipeline_tag, tags
Browse files
README.md
CHANGED
|
@@ -7,9 +7,31 @@ tags:
|
|
| 7 |
- video generation
|
| 8 |
- video-to-video editing
|
| 9 |
- refernce-to-video
|
| 10 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
pipeline_tag: image-to-video
|
|
|
|
|
|
|
| 12 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
# Wan2.1
|
| 14 |
|
| 15 |
<p align="center">
|
|
|
|
| 7 |
- video generation
|
| 8 |
- video-to-video editing
|
| 9 |
- refernce-to-video
|
| 10 |
+
- wan
|
| 11 |
+
- wan2.1
|
| 12 |
+
- vace
|
| 13 |
+
- image-to-video
|
| 14 |
+
- video-to-video
|
| 15 |
+
- video-generation
|
| 16 |
+
- bf16
|
| 17 |
pipeline_tag: image-to-video
|
| 18 |
+
base_model: Wan-AI/Wan2.1-VACE-14B
|
| 19 |
+
base_model_relation: quantized
|
| 20 |
---
|
| 21 |
+
## About this repository
|
| 22 |
+
|
| 23 |
+
This is a **bf16 cast** of [Wan-AI/Wan2.1-VACE-14B](https://huggingface.co/Wan-AI/Wan2.1-VACE-14B),
|
| 24 |
+
published by WaveSpeed AI. Upstream stores the diffusion transformer in fp32;
|
| 25 |
+
here it is bf16, which halves the download from ~69.4 GB to ~34.7 GB with no
|
| 26 |
+
change to tensor names or shapes. Nothing is quantized.
|
| 27 |
+
|
| 28 |
+
The UMT5 text encoder (`models_t5_umt5-xxl-enc-bf16.pth`) and the VAE
|
| 29 |
+
(`Wan2.1_VAE.pth`) are carried over from upstream unchanged.
|
| 30 |
+
|
| 31 |
+
The rest of this card is Wan-AI's original VACE model card, kept for reference.
|
| 32 |
+
|
| 33 |
+
---
|
| 34 |
+
|
| 35 |
# Wan2.1
|
| 36 |
|
| 37 |
<p align="center">
|