SynthForensics: Benchmark for People-Centric Synthetic Video Deepfakes

SynthForensics is a large-scale benchmark dataset for synthetic video forgery detection. It comprises fake videos produced by multiple Text-to-Video and Image-to-Video generators, organized to align with the canonical FaceForensics++ protocol. The dataset was created by SynthForensics and was last updated on 2026-04 30.

Use Cases

• Benchmarking the generalization of deepfake detectors based on the dataset's alignment with the FaceForensics++ protocol.
• Evaluating detection performance against Text-to-Video (T2V) generators based on the described fake video sources.
• Evaluating detection performance against Image-to-Video (I2V) generators based on the described fake video sources.
• Training forgery detection models on a large-scale collection of synthetic videos.

Strengths

• Designed as a large-scale benchmark, suggesting substantial volume.
• Includes fake videos from multiple state-of-the-art generative models (T2V and I2V).
• Organized to align with the canonical FaceForensics++ (FF++) protocol for comparability.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file size are unknown, which may limit suitability assessment.

Provenance

Source

huggingface

Collection Method

Fake videos produced by multiple Text-to-Video (T2V) and Image-to-Video (I2V) generators.

Time Range

null

Freshness

Last updated 2026-04-30 15:42:28; freshness should be verified.

Geography

null