SD Exploration Round 1

[chunxi]

AI Output

Tested some prompts with AnimateDiff and ModelSpace

AnimateDiff

1. prompt "((best quality)), ((masterpiece)), ((realistic)), long highlighted hair, cybergirl, futuristic silver armor suit, confident stance, high-resolution, living room, smiling, head tilted"

2. prompt "((best quality)), ((masterpiece)), ((realistic)), long highlighted hair, cybergirl, futuristic silver armor suit, confident stance, high-resolution, medieval castle, winking"

ModelSpace

1. prompt "a young woman in a flowy dress winks left eye" https://github.com/chunxi/ai-video-playground/assets/1466088/a4410a34-b6c8-4b97-b75f-f40f9ea9d9b7

2. prompt "a young woman, winking left eye, looking at viewer, smiling, beautiful symmetrical face" https://github.com/chunxi/ai-video-playground/assets/1466088/03858895-ca3d-4b66-8487-c0c7cba6cfd4

3. prompt "a young woman, winking left eye, looking at viewer, smiling, beautiful symmetrical face, mouth, noise, neck, upper body." https://github.com/chunxi/ai-video-playground/assets/1466088/0627bba6-c01f-4c6c-9231-707f866665f5

Comparison

• AnimateDiff and img2vid both generates an image with the prompt. Prompt has no control over how the video is generated.
• ModelSpace's video is controlled by the prompt. But which prompt it takes in is less predictable. And the video is very close up.