Skip to content

Dataset · APRIL-AIGC

Multimodal benchmark for avatar animation

APRIL-AIGC/Soul-Bench

Dataset of up to 1 thousand examples for evaluating image-to-video generation and digital human animation.

What's inside

The benchmark includes fewer than 1 thousand examples and combines four modalities: video, images, audio, and text. The set was created by the Soul project authors to evaluate long-term multimodal animation of digital avatars and solve image-to-video tasks.

What you can do with it

  • Evaluating and comparing image-to-video generation models
  • Testing synchronization systems for speech, facial expressions, and movements of digital avatars
Tasks
Image → video
Size
up to 1K rows
Downloads this month
730
Likes
56

How to load it

from datasets import load_dataset

ds = load_dataset("APRIL-AIGC/Soul-Bench")