Douyin Viral Short Video Decomposition

Pain Points in Douyin Video Analysis

Manually decomposing viral Douyin videos is time-consuming and lacks systematization. Content creators need to quickly extract structural segmentation, viral attribution, and reusable strategies from competitor videos, but traditional methods rely on manual viewing and note-taking, which is inefficient and prone to missing key details.

How the Skill Works

This skill automates video decomposition via an asynchronous interface. Core capabilities and key steps include:

• Input Handling: Accepts Douyin share links (short https://v.douyin.com/... or long with /video/) or local video files (e.g., .mp4, .mov). Local files are automatically uploaded to the server and converted to a video_id.
• Task Scheduling: Uses python3 scripts/analyze_dy_video.py to submit tasks.
• --only-create creates a task and returns an analysis_task_id.
• --poll-task polls for progress to avoid blocking the session. Tasks run for 1–5 minutes, requiring multiple short polls (each ≤90 seconds).
• Report Generation: On success, extracts a Markdown report from data.result.markdown, detailing video structure segmentation, viral cause analysis, scoring, and strategy recommendations. The report is rendered in Markdown format, including headings, tables, and lists.

Key steps like point deduction confirmation (about 128 points) and exit code handling (e.g., exit code 13 indicates in-progress status requiring continued polling) ensure process reliability.

Applicable Boundaries and Notes

• Input Limitations: Only supports Douyin platform videos; other platform links (e.g., WeChat Channels, Bilibili) require prior download as local files. File formats are limited to mp4/mov etc., with a size cap of 200MB—compression is needed if exceeded.
• Point Mechanism: New tasks deduct points, settled upon success; points are refunded on failure. Polling or retrying the same task (--retry-task) does not incur additional deductions, but true retries require re-initiation.
• Report Accuracy: Decomposition is generated by remote services for reference only; critical decisions should involve human review. Report Markdown must be rendered correctly to avoid formatting errors.
• Interaction Protocols: Steps must be followed sequentially: confirm API Key (configured in config.json), recognize input, confirm point deduction before execution, and finally deliver the report by rendering the Markdown content.