Take one Douyin video apart: real plays, the spoken script, how a placement was worked in, and what the comments say.
v0.1.1
View example draft · How to useExample draft
A 2 min 15 s recipe video by 周幺姑家常菜 with 241,691 plays and a seasoning placement.
The product enters as the answer to a stated problem (people do not know how to season ribs) and replaces several steps, so it does not interrupt the recipe. Concrete quantities (180 g, 2.5–3 jin, 20 min, 35 min) make the video worth saving, which matches the high collect share (33% of likes).
1. Hook: "<dish> 比 <method A> 简单，比 <method B> 好吃" 2. Credibility: ingredient choice with real prices 3. Prep step with one non-obvious tip 4. Pain point → product as the shortcut, with a quantity rule 5. Cook with exact times; garnish and payoff
Excerpts from an actual run on 2026-11-05.
An Agent that can read the complete Skill folder. Live research also needs a SandBase connection; supplied material can be used without a lookup. The transcript uses a paid speech-to-text model through SandBase.
Download and unpack the complete Skill folder, keep its references, and use it with a local Agent that supports Skills.
Use douyin-video-breakdown on this Douyin video. Give me the real play count, the transcript and its structure, how the product placement is introduced, what the comments say, and a script skeleton I can reuse.
Read the detail and the statistics call, which carries the real play count.
Run speech-to-text on the original audio track and mark doubtful dialect words.
Label hook, steps, placement and close; tag purchase intent and questions in the comments.
Download and unpack the whole Skill folder, including references, into your Agent’s supported Skill directory. You can also ask the Agent to read its SKILL.md and follow the instructions. Run it locally with your own material.
No. It is machine speech recognition; dialect speech produces wrong characters, which the report marks.