📢 Project Report: 240919
주요 기능
- 동요 스타일 비디오 생성을 위한 프로젝트
- Text-to-Video or Image-to-Video 관련 딥러닝 코드 구현 및 실험
- AnimateDiff, CogVideo, ControlVideo, Free-Bloom 모델 구현 및 실험
View Project & Code
1. Prompts List
The following prompts were used to generate videos with different styles using various models:
-
Realistic Style:
“A video of a duckling wearing a medieval soldier helmet and riding a skateboard.” -
Cartoon Style:
“A video of a duckling wearing a medieval soldier helmet and riding a skateboard in cartoon style.” -
Watercolor Style:
“A video of a duckling wearing a medieval soldier helmet and riding a skateboard in watercolor style.”
2. Comparison Models
The video generation was performed using the following models:
- AnimateDiff
- Free-Bloom
- CogVideo (available on Hugging Face)
- ControlVideo
3. Inference
You can run the inference code for each model from the following notebook:
#Start path
git clone https://github.com/ssoojeong/Baby_Video.git
cd Baby_Video
- Notebook Path:
./inference_code/inference.ipynb
This notebook is configured to support running inference for comparison models.
4. Results
You can see the results of the inference for each model from the following notebook:
- Notebook Path:
./inference_outputs/{model_name}
View Sample Results
