Skip to content

wan_prompt_expand

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History
 
 

wan_prompt_expand

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
LICENSE
LICENSE
 
 
README.md
README.md
 
 
__init__.py
__init__.py
 
 
prompt_expand.py
prompt_expand.py
 
 
qwen_vl_utils.py
qwen_vl_utils.py
 
 

README.md

WAN 2.1 Prompt Expansion

Expands short prompts into detailed descriptions optimized for WAN video generation using Qwen language models.

Usage in Inference

Enable prompt expansion via command-line arguments in scripts/inference/video_model_inference.py:

python scripts/inference/video_model_inference.py \
    --prompt_expand_model Qwen2.5_14B \
    --prompt_expand_model_seed 42 \
    ...

Arguments:

  • --prompt_expand_model: Qwen model to use (Qwen2.5_3B, Qwen2.5_7B, or Qwen2.5_14B)
  • --prompt_expand_model_seed: Random seed for reproducible expansion (default: 0)

When enabled, prompts are expanded on rank 0 and broadcast to all ranks. The expanded prompt includes subject details, visual style, camera angles, and motion attributes.

Source

Code from Wan-Video/Wan2.1 at commit 7c81b2f. Extracted prompt expansion from wan/utils and removed dashscope dependency.

License

See licenses/Wan/LICENSE for details.