TubeLab · Video script
TubeLab · Write scripts for different video channels and compare story organisation, style, pacing and clarity.
What it measures, and how
Unifies TubeLab Scriptwriter, 12 tasks and 6 writing-quality measures; three model referees do not judge their own family, and failed tasks keep their penalty. Cost and speed are not part of ability.
How this evidence is used
Under observation: awaiting stable raw data, an accurate evaluation date and reuse boundaries, rather than using a display page in place of verifiable scores.
Limits and data attribution
Models are affected by the Scriptwriter flow; OpenAI uses the lowest reasoning tier, which cannot be extrapolated to other tiers. Spoken video is not the same as a film screenplay.
Data licence: Stable score export and public aggregation permission are unconfirmed.
Scores published by TubeLab. Raw scores and the News consensus score use different scales and cannot be added directly.