What problem does it solve?
Editing Chinese talking-head videos requires finding filler words, repeated takes, misspoken names, and false starts across long transcripts, which is slow and error-prone when done manually. This Skill turns a raw video into a reviewed deletion ledger (Cuts + EDL) without physically cutting any media.
Core Features & Use Cases
- Word-level transcription and dictionary correction: Cloud ASR transcribes the video word-by-word, then general and user dictionaries fix misheard proper nouns before any deletion judgment.
- Five-round semantic deletion scan: Each pass looks for exactly one category—repetition, broken sentences, misspoke-and-retry, English stumbles, and filler words—producing a summary table and a repeated-sentences table.
- Human review in Studio with retrospective learning: Candidates are submitted with revision-safe CAS semantics, the user reviews in the product Studio, and corrections are archived into user preference and dictionary files.
- Use Case: A creator records a 30-minute Chinese commentary video with multiple retakes. The Skill transcribes it, fixes misheard terms like Grok or Codex, proposes deletions for repeated takes, and hands a reviewed cut list to the export stage.
Quick Start
Ask the agent to cut this talking-head video by transcribing it, fixing words with the dictionary, finding misspeaks and repetitions, and opening Studio for review.
