Audio2Face Mediapipe Arkit Teacher: 14,703 Emotional Speech Clips with ARKit Blendshapes

14,703 emotional-speech clips annotated with 52-channel ARKit blendshape sequences extracted by MediaPipe FaceLandmarker from source video or audio-driven synthesis. The dataset was created by myned-ai and was last updated on 2026-06-11.

Use Cases

• Train models for audio-to-facial-animation based on the 52-channel ARKit blendshape vector.
• Benchmark facial landmark detection and tracking algorithms using the MediaPipe FaceLandmarker outputs.
• Develop emotional speech synthesis systems that drive realistic facial expressions.
• Create tools for generating ARKit-compatible facial rigs from audio or video input.

Strengths

• Contains 14,703 emotional-speech clips, providing a substantial collection for training.
• Each clip is annotated with a 52-channel ARKit blendshape sequence, offering detailed facial motion data.
• Data was processed using MediaPipe FaceLandmarker, a standardized computer vision pipeline.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known, but other scale metrics like file formats and size are unknown.
• Freshness should be verified as the last update timestamp is in the future (2026-06-11).

Provenance

Source

myned-ai

Collection Method

ARKit blendshape sequences extracted by MediaPipe FaceLandmarker from source video or audio-driven synthesis.

Freshness

Last updated 2026-06-11 11:24:14