AI Video Generator

Generate high-quality AI videos from text prompts or images using state-of-the-art open-source models.

Models Supported

Model	Type	Quality	Length	Best For
CogVideoX-2b	Text-to-Video	720p	6 sec	Best overall quality
CogVideoX-5b	Text-to-Video	1080p	6 sec	Premium quality (needs 26GB VRAM)
SVD-XT	Image-to-Video	1024x576	4 sec	Animate images
Open-Sora	Text-to-Video	768p	5 sec	Full framework
Zeroscope	Text-to-Video	1024x576	3 sec	Fast generation

Usage

Generate Video from Text

python3 ai_video.py text "A cat walking on the moon, cinematic lighting, 4k"

Generate Video from Image

python3 ai_video.py image /path/to/image.jpg

Python API

from ai_video import generate_video_cogvideox, generate_video_svd

# Text to video
video_path = generate_video_cogvideox(
    prompt="A beautiful sunset over mountains, cinematic",
    num_frames=49,
    fps=8,
    output_name="sunset.mp4"
)

# Image to video
video_path = generate_video_svd(
    image_path="input.jpg",
    motion_bucket_id=127,
    fps=7,
    output_name="animated.mp4"
)

Hardware Requirements

Model	Min VRAM	Recommended
CogVideoX-2b	4GB	8GB+
CogVideoX-5b	16GB	24GB+
SVD-XT	8GB	12GB+
Open-Sora	16GB	24GB+

Tips

• Use descriptive prompts for better results
• Keep videos short (3-6 seconds) for best quality
• Use negative prompts to avoid artifacts
• Enable CPU offload for lower VRAM usage