Source: the first 20 seconds of 4 highly liked videos from Douyin "小Lin说", breakdown on 2026-10-27. Samples see
references/samples.md. This Skill only distills the actions, rhythm, and structure; it does not include the other party's appearance, clothing, scene, or verbatim lines. When generating, do not feed the original video to the model as a reference video.
In one sentence
At the start, show the source material first, with the person in a small window in the lower-left corner doing a voiceover and telling the story; only on the most emotional sentence do you return to full screen, with the hands making a "shaped" big gesture along with it, as if gesturing an object in the air; when reasoning, drop to a single hand.
Suitable / Not suitable
- Suitable: finance, business, hot-topic analysis, storytelling-style openings about people and phenomena
- Not suitable: reference images where the hands are not in frame; digital human projects with high requirements for hand stability
- Digital human difficulty:High. Finger shapes are complex.The structure can be borrowed as a whole set; for gestures, it is recommended to borrow only X4
Camera position and composition
- Seated, medium close-up from the chest up, person centered, both hands can be raised into frame
- Background: a tidy bookshelf or a corner of a living room with a personal logo light sign, warm tones. Describe according to your own scene
- Fixed camera position
Gesture vocabulary
| Code | Gesture | Used for which type of line | Position and amplitude | Duration |
|---|---|---|---|---|
| X1 | Both hands with five fingers spread forcefully, on both sides of the face, palms facing the camera or inward, with a slight shake | The most emotional words: "finally", "blockbuster", the annual theme name | Shoulder-to-cheek height, beyond shoulder width | 1–2 seconds |
| X2 | Thumb and index finger pinched together, in front of the chest, tapping with the rhythm | Word-by-word emphasis | In front of the chest, slightly to one side | About 0.25 seconds per tap |
| X3 | Ten fingers bent as if holding a ball, fingertips facing each other, fingers opening and closing with the syllables | Describing visual imagery: "another world", "dazzling" | Chest to chin | 1–3 seconds |
| X4 | One hand held as a side palm at one side of the body, palm facing inward, tracing a small arc with the sentence | Adding background, reasoning: "You should know..." | Side of the chest | 0.5–1 second |
Rhythm rules
- The first sentence back on full screen is usually the most emotional sentence, and the gesture is also the largest (X1, X3)
- When reasoning, drop to a single hand (X4)
- Gestures land on the stressed syllables; two-handed gestures are symmetrical
Expression and head
- Large expression amplitude: wide eyes, open mouth, raised eyebrows, body leaning forward, exaggerated mouth shapes
- Smile immediately after finishing the punchline
Opening structure template
Regular video:
| Time | Visual | Action | Dialogue function |
|---|---|---|---|
| 0–9s | News photos and live-action footage full screen, person in a small circular window at the lower left | Only facial expressions in the small window | Introduce a person or a phenomenon, tell the story in colloquial speech |
| 9–12s | Cut back to the real person full screen | X1 or X3 + exaggerated expression | Emotional sentence: contrast, exclamation |
| 12–18s | Cut back to the footage | — | Key facts: numbers, events |
| 18–20s | Real person full screen | X4 one-hand side palm | “You should know…” to point out why it matters |
Annual or major video: Real person full screen with an emotionally charged opening line + X1 → title card → fast-cut news footage.
Write the dialogue yourself; only learn the function of the sentences.
Digital human generation suggestions
- The whole structure is usable: cold open on footage + voiceover in the small window; the digital human appears full screen only 2 times, 2–4 seconds each
- Use only X4 for gestures; if you must express “maximum emotion,” use an expression plus leaning forward instead of X1, X3
- The person in the small window can be generated at low resolution or as a still frame with lip-sync to reduce cost
- The footage segments are done at the editing layer, which conveniently covers the seams
Prohibited actions
- The digital human does not do finger close-ups in X1, X2, X3 (unless the model has proven it can generate fingers reliably)
- Hands do not go above the head or extend beyond either side of the frame
- No head shaking, no shrugging
Output format
## 基础姿态
## 逐句动作
| 时间 | 台词 | 画面（素材+小窗 / 真人全屏） | 手势代号与描述 | 表情与头部 |
|---|---|---|---|---|
## 禁止动作
## 给视频模型的描述（只写真人全屏段）
## 素材清单（冷开场和事实段要用的画面）
Generation description example (original, may be rewritten)
Medium close-up, the person sits facing the camera, framed from the chest up. The camera is static. At second 0 the person opens their eyes wide, raises their eyebrows, leans forward, and speaks with exaggerated lip movements. At second 1.5 the right hand rises on the right side of the body, palm facing inward, traces a small arc with the sentence, then lowers. At the end of the sentence the person smiles with teeth showing. The left hand does not enter the frame.