Incantation Elden Ring: Manually Collected Combat Clips with Structured Captions

A preview subset of the Incantation dataset, released in May 2026 by MatrixTeam. It contains manually collected Elden Ring combat video clips paired with structured captions, documenting the data format and annotation style ahead of a full project release. The dataset is described as a reference subset and not the final, complete benchmark.

Use Cases

• Training video-to-text models based on structured captions for combat scenes.
• Benchmarking action recognition algorithms on manually annotated game footage.
• Developing game AI agents by learning from structured descriptions of in-game combat.
• Researching multimodal annotation styles for complex, dynamic video content.

Strengths

• Data is manually collected, suggesting a degree of curation and quality control.
• Clips are paired with structured captions, providing a foundation for supervised learning.
• Released by MatrixTeam, indicating an organized effort behind the data creation.

Limitations

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

Provenance

Source

MatrixTeam via Hugging Face.

Collection Method

Manually collected Elden Ring combat clips.

Freshness

Last updated 2026-05-26 13:11:01; freshness should be verified.