Douyin Video Copy Extraction

Problem

When reviewing Douyin videos, spoken content is often buried in the video itself. Manual transcription, cleanup, and reformatting are costly, and easily drop the original phrasing. This skill turns the task into a structured workflow: provide a video URL, shared text, or aweme_id, then receive video metadata, the original description, spoken transcript, copy-ready text, and a concise summary.

How It Works

• Input support: accepts --url, --aweme-id, or an existing job_id for follow-up polling.
• Async polling: submit the task first, check data.is_terminal; if it is not terminal, keep querying the same data.job_id using data.next_poll_after_seconds when available.
• Layered output: after success, prioritize description, transcript, copy-ready version, and summary without inventing fields that were not returned.
• Failure handling: suggest retry only when data.error.retryable is true; handle insufficient balance according to the returned error instead of resubmitting.

Boundaries

This skill is focused on read-only transcript extraction. It does not log in, post, comment, like, or modify account state. Unless explicitly requested, it should not expand into rewriting, ad analysis, or content from other platforms.