trentmkelly/disturban-scripts

Disturban scripts

This dataset contains scripts from the YouTuber Disturban.

The records were parsed with Gemini 3 Flash to generate both voiceover text and video descriptions, so this dataset is not just audio transcripts. It is structured for workflows that need paired narration and visual-scene descriptions.

Files

• —disturban.jsonl: JSONL records containing parsed Disturban script content and generated descriptions.

Possible uses

• —Narration-to-video or script-to-shot-description modeling
• —Video planning and storyboarding experiments
• —Long-form script analysis
• —True-crime content structure analysis

Content warning

Disturban content often covers crime, violence, disturbing events, and other sensitive topics. Downstream users should expect potentially graphic or distressing subject matter.