Source: the first 20 seconds of 4 highly liked videos from Douyin "Qiu Zhi 2046", analyzed on 2026-10-27. See samples at
references/samples.md. This Skill only distills the movements, rhythm, and structure; it does not include the person's appearance, clothing, setting, or verbatim lines. When generating, do not feed the original videos to the model as reference videos.
In one sentence
The hands almost never leave the frame, switching gestures every 0.5–1 second on average, but it is only ever the same 6–7 fixed gestures, all two-handed and symmetrical, elbows tucked in, landing on the stressed syllables. It looks busy but is actually very disciplined.
Suitable / not suitable
- Suitable: AI tool tutorials, beginner introductions, openings that need approachability and energy
- Not suitable: serious opinions, financial analysis; digital humans whose hands are not in frame in the reference image (switch to
talking-head-li-yifan） - Digital human difficulty:High. Many fingers and fast switching make hands the most likely to break. It is recommended to use only 3–4 of these gestures
Camera position and composition
- Seated, medium close-up from the chest up, subject centered, a little headroom above the head
- The full hand movement area is in frame (the bottom edge cuts slightly above the waistline)
- Fixed camera, no push-in or pull-out; let editing handle shot-size changes (cut to picture-in-picture)
- Background: a studio with a color cast, with one neon or lighting element, background blurred. Describe it according to your own scene; do not copy it directly
Gesture vocabulary
| Code | Gesture | Used for which type of line | Position and range | Duration |
|---|---|---|---|---|
| Q1 | Both hands open, palms facing the camera, wrists gently swaying left and right | Opening greeting (present in all 4 samples) | From shoulder to both sides of the cheeks, slightly wider than the shoulders | 0.75–1 second |
| Q2 | Both hands with index fingers pointing up, gently tapping up and down with the syllables | "This is an episode of…", naming a key term or number, calling for follows | Chin to eyebrow height, close to both sides of the face | 1–2 seconds, one tap per stressed syllable |
| Q3 | Both hands in hollow fists, lifted or gently swaying in front of the chest | Call to action, wrap-up: "Let's start", "Give us a follow" | Chest to chin, elbows tucked in | About 1 second |
| Q4 | One hand with an upright palm, the other in a fist | Promise line: "After watching, you'll be able to compare…" | Upright palm at the side of the face, fist in front of the shoulder | 1–1.5 seconds |
| Q5 | Both palms spread from the chest to the sides or draw a half arc | Expansion lines like "could all be one of your essential skills" | No more than 1.5 times the shoulder width | About 1 second |
| Q6 | Both hands pressed together in front of the chest | Transitional filler: "Hehe, today…" | Center of the chest | 0.5–1 second |
| Q7 | Both hands extended forward, palms facing down | Closing 「出发」「Let's go」 | extend toward the lower part of the frame | about 0.5 seconds |
Beat rules
- Gestures switch only on line stresses; one stress corresponds to one gesture change or one tap
- Mainly symmetric with both hands; only Q4 is asymmetric
- Hands always stay within the 「chest to eyebrows」 frame: not below the tabletop, not above the top of the head
- Elbows stay close to the body; movements are done with the forearms and wrists
Expression and head
- Laughing with teeth showing throughout
- Squint or close the eyes briefly on stressed words (about 0.3 seconds), as if stamping that word
- Head tilts slightly left and right, nodding lightly with the sentence; gaze stays on the camera
Opening structure template
| Time | Visual | Action | Line function |
|---|---|---|---|
| 0–1s | Full-screen live person | Q1 wave | Greeting |
| 1–6s | Full-screen live person | Q2 double index fingers, Q4 palm plus fist | One-line hook: what you can learn in this episode, and what you can do after learning it |
| 6–16s | Picture-in-picture: person in a vertical frame on the left, table of contents or demo on the right | Facial expressions dominate in the small window | List this episode's content |
| 16–18s | Back to full screen | Q2 or Q3 | Ask for saves and follows |
| 18–20s | Full-screen live person | Q3 or Q7 | 「我们开始」, enter the main content |
Write the lines yourself. The original author has fixed greeting phrases and a closing slogan; learn only the function of the sentences, do not copy them.
Digital human generation suggestions
- Choose only 3–4 from the vocabulary list: Q1 (opening), Q2 (hook), Q3 (call to action) are recommended, plus at most Q5
- Place only 1–2 gestures per generated segment, and leave ≥1.5 seconds of 「hands lowered, just talking」 between the two gestures
- Do the 6–16 second picture-in-picture segments at the editing layer, exactly covering the seams between two generated segments
- The reference image must be a medium close-up with both hands in frame; if the hands are not in frame in the reference image, switch to an expression-led style
Prohibited actions
- Hands do not go above the top of the head and do not extend beyond the left or right edges of the frame
- Do not point at the camera with one finger
- Do not cross fingers; do not make complex hand shapes such as finger hearts, OK signs, or counting numbers
- Do not change gestures more than 3 times in a row
- Do not shake the head, do not shrug, do not sway the body left and right
Output format
After receiving the user's lines (and optional reference image), output the action specification in this format:
## 基础姿态
（机位、景别、坐姿、手的默认位置）
## 逐句动作
| 时间 | 台词 | 画面 | 手势代号与描述 | 表情与头部 |
|---|---|---|---|---|
## 禁止动作
（照抄上面的列表，按需补充）
## 给视频模型的描述
（每个生成段一段，写清楚：人物姿态、手势的开始时刻与位置、表情、镜头不动）
Generation description example (original, may be rewritten)
Medium close-up, the person sits directly facing the camera, framed from the chest up, with both hands in frame. Camera is fixed. At second 0, the person laughs with teeth showing, raises both hands to the sides of the shoulders with palms facing the camera, gently sways the wrists left and right twice, then places them in front of the chest. From second 1.5, both hands raise their index fingers to the sides of the chin, tapping up and down once each with the stresses of speech, squinting briefly with each tap. Hands always stay between the chest and the eyebrows, elbows close to the body, and no other gestures are made.