DigiFakeAV: Synthetic and Real Video Clips with Audio

The DigiFakeAV dataset contains synthetic and real video clips, with video frames stored as raw numpy arrays and audio provided in uncompressed WAV format. It was created by author 'cambrain' and last updated on March 12, 2026. The dataset is part of a project focused on audio-visual media.

Use Cases

• Training deepfake detection models based on the mix of synthetic and real video clips.
• Developing multimodal AI systems based on the separate video and audio file structure.
• Benchmarking audio-visual forgery detection algorithms based on the provided video frames and audio samples.

Strengths

• Provides separate, uncompressed audio files (.wav) for clear audio analysis.
• Video frames are stored as raw numpy arrays with accompanying metadata files for dimensions and frame count.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

cambrain

Collection Method

Part of the DigiFakeAV project; likely contains both generated and collected media.

Time Range

null

Freshness

Last updated 2026-03-12 02:46:54; freshness should be verified.

Geography

null