VideoASMR-Bench: 100 AI-Generated ASMR Videos for Realism Evaluation

VideoASMR-Bench is a benchmark for evaluating the realism of AI-generated videos, specifically focusing on ASMR content. The dataset contains 100 test samples across two difficulty levels, created by author kolerk. It was last updated on May 5, 2026.

Use Cases

• Benchmarking video generation model realism based on ASMR-specific criteria like texture rendering.
• Evaluating audio-visual synchronization in generated content as described in the benchmark focus.
• Testing visual language model (VLM) perception of AI-generated video realism.
• Conducting human evaluation studies on the perceived authenticity of synthetic ASMR videos.

Strengths

• Contains 100 test samples explicitly designed for evaluation.
• Structured with two difficulty levels to provide a tiered assessment challenge.
• Focuses on the specific, demanding domain of ASMR, which requires high fidelity in multiple sensory dimensions.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown beyond the 100 test samples, which may limit suitability assessment for large-scale training.
• Description metadata is limited; actual data quality and file formats require manual inspection after download.

Provenance

Source

huggingface

Freshness

Last updated 2026-05-05 05:41:51; freshness should be verified.