成片拆解
来源 · X

棕色乐福鞋 · 开箱试穿种草 · Seedance 2.5
23 秒真人风短片,怎么做出来的?

棕色乐福鞋开箱种草:拆盒、看细节、试穿、走到镜头前,手机手持拍法,配口播和大字幕。

套路 · 产品广告套路 · 手机POV·Vlog23秒 · 开箱种草9:16 竖屏10 个镜头商品四视图参考Seedance 2.5

钩子在哪

时间为抽帧估计

结构:抱盒开口 → 拆盒 → 手持看鞋 → 微距看细节 → 穿上 → 俯视看脚 → 走到镜头前

看完成片再对照:这条片子靠什么在几秒内抓住人。点时间可跳到成片对应位置。

开场钩子

第 0 秒她坐在床边,把米色鞋盒捧到胸前,对着镜头说「Okay, these finally came」。

拆盒和看鞋

约 3–6.3s 俯拍掀盖、再切到她坐在地板上翻开白色包装纸;6.5–8.3s 一只手把鞋举起,四分之三侧面对镜头。

细节

约 8.5–10.8s 微距,指尖沿着鞋面横带和冲孔滑过;11–12.3s 翻过来看黑色锯齿大底;12.5–14.8s 双手拿鞋举到镜头前。

上脚和收尾

约 15–17.3s 低机位看脚伸进鞋里踩实;17.5–19.8s 站着往下拍两只鞋;20–23.4s 低机位全身,她从左边走进来站定面对镜头。

跟做照抄这一点

把鞋的四个视角分别当参考图,并写清每张图对应哪个角度;每个镜头都写机位、焦段、手在哪、光从哪来,让手持手机的质感贯穿全片。

三步搞懂

别急着上视频模型。这个片子的秘密是:先把角色和场景锁死,再一次生成 10 个镜头。

1

第一步:准备人物和鞋子参考图

作者的流程:用 Recraft 做人物,用 GPT-2 做鞋子映射,Seedance 2.5 生成视频,ElevenLabs 配音,CapCut 剪辑。回复区给了三张图:人物脸部特写、提示词长图、一张鞋子四视图(正面、四分之三侧、俯视、正侧面)。人物特写原图收录;鞋子四视图裁成四张收录。

2

第二步:锁定鞋子,写全局设定

REFS 写明每张鞋图对应的角度,要求全片是同一双鞋:鞋头形状、带钥匙孔和冲孔的横带、皮纹、中棕色和擦色边、黑色锯齿大底;棚拍背景和灯光不要带进画面。全局设定写:左边一扇窗的上午阳光、手机视频质感、卧室橡木地板、人物服装(米白罗纹毛衣、卷边靛蓝牛仔裤、米白袜、右手两枚银戒),脸只在第 1 和第 8 镜出现。

3

第三步:按 8 个镜头写开箱到上脚

8 个镜头共 30 秒:抱盒开口 → 俯拍掀盖 → 翻开包装纸取鞋 → 侧面转鞋 → 横带微距 → 脚伸进鞋 → 俯视双脚 → 低机位看她走进画面。每镜写景别、角度、焦段、手持方式、动作和声音;只有第 1 镜有一句台词,最后要求不要字幕、不要背景音乐。

步骤一 · 做出这 5 张参考图

推荐用 GPT Image / 同等图模,比例 4:3。每张点开即可复制完整英文提示词。

5

@[Image 5] · 人物脸部特写

作者回复区第一张图原件 HQvO0dDWcAIWFHx(896x1216)· 作者说人物用 Recraft 做

第 1 步
@[Image 5] · 人物脸部特写
▸查看完整出图提示词(点击复制区)
复制
作者未公开出图提示词(主帖写人物用 Recraft 生成)。视频提示词中的用法:REFS: @[Image 5]girl。编号对应是按提示词推断,作者没有单独标注。
1

乐福鞋 · 正面

提示词里对应 @[Image 4]front;回复区商品四视图按四等分裁出

第 2 步
乐福鞋 · 正面
▸查看完整出图提示词(点击复制区)
复制
作者说鞋图用 GPT-2 做了映射,没有公开出图提示词。
2

乐福鞋 · 四分之三侧

提示词里对应 @[Image 1](three quarter;回复区商品四视图按四等分裁出

第 3 步
乐福鞋 · 四分之三侧
▸查看完整出图提示词(点击复制区)
复制
作者说鞋图用 GPT-2 做了映射,没有公开出图提示词。
3

乐福鞋 · 俯视

提示词里写作 @overhead,没有编号;回复区商品四视图按四等分裁出

第 4 步
乐福鞋 · 俯视
▸查看完整出图提示词(点击复制区)
复制
作者说鞋图用 GPT-2 做了映射,没有公开出图提示词。
4

乐福鞋 · 正侧面

提示词里对应 @[Image 2]profile;回复区商品四视图按四等分裁出

第 5 步
乐福鞋 · 正侧面
▸查看完整出图提示词(点击复制区)
复制
作者说鞋图用 GPT-2 做了映射,没有公开出图提示词。

步骤二 · 10 镜头故事板(人话版)

方便你对照成片理解每一镜在干什么。

  1. Shot 1 — 0–3s 坐在床边,捧着鞋盒对镜头说第一句。
  2. Shot 2 — 3–4.3s 俯拍,戴银戒的手掀开盒盖,露出白色包装纸。
  3. Shot 3 — 4.5–6.3s 她坐在地板上,双手翻开包装纸。
  4. Shot 4 — 6.5–8.3s 手把一只鞋举起,四分之三侧面对镜头。
  5. Shot 5 — 8.5–10.8s 微距,指尖沿鞋面横带和冲孔滑过。
  6. Shot 6 — 11–12.3s 把鞋翻过来,展示黑色锯齿大底。
  7. Shot 7 — 12.5–14.8s 站着双手拿鞋,举到镜头前。
  8. Shot 8 — 15–17.3s 低机位,穿米白袜的脚伸进鞋里踩实。
  9. Shot 9 — 17.5–19.8s 站着往下拍两只鞋。
  10. Shot 10 — 20–23.4s 低机位全身,她从左边走进画面站定。
关键约束:同一双鞋贯穿全片;人物服装不变;脸只在开头和结尾出现;不要字幕、不要背景音乐。与成片不符:提示词是 8 镜 30 秒,成片 23 秒、约 10 个镜头(多了坐在地板上拆纸、看鞋底、举鞋到镜头前);提示词只有第 1 镜一句台词,成片全程有口播(作者说配音用 ElevenLabs);提示词写不要字幕,成片有逐词弹出的大字幕,应是 CapCut 后期加的;提示词写深棕中分及肩直发,成片是及肩以下的卷发,和人物参考图一致。原文 SHOT 3 里「Tissue cracklre」是原图拼写,照录。缺口:@[Image 5] 对应脸部特写是按提示词推断,作者没有说明编号对应;口播台词没有逐句核对。

步骤三 · 完整视频提示词

直接全选复制到 Seedance。@Image 1–5 必须对应上面 REF 01–05。

GEN 01

Unboxing → Reveal → Wear → Walk

Seedance 2.5 · 9:16 · 英文完整提示词(作者回复区第二张图,逐字转录)

复制
UNBOXING → REVEAL → WEAR → WALK
SHOT STRUCTURE: 8 shots, 30 seconds, 9:16, exactly as listed, no added shots

REFS: @[Image 5]girl
@[Image 4]front, @overhead, @[Image 1](three quarter, @[Image 2]profile. One pair of loafers, the same pair in every shot. They control toe shape, penny strap with keyhole and perforated dots, leather grain, mid brown with burnished edges, black lug sole and tread. Nothing else: the studio background and studio lighting stay out, and every shot builds fresh framing of its own.

ACTIVE REFERENCES: @[Image 4](image_4) to @[Image 1](image_1) all 100%

GLOBAL STYLE NOTES:
- lighting_philosophy: one window screen left, late morning sun through thin curtain, warm bounce off the oak floor
- color_grade: modern phone video, faint sensor noise in shadows. ~60% warm neutral (linen, oak, wall) + ~30% mid brown (leather) + ~10% indigo (denim).
- setting: bright bedroom, wide oak floorboards, unmade linen bed, matte sand shoe box
- ambience: quiet room tone, muffled street through glass
- characters: <Girl> cream ribbed knit sweater, sleeves at mid forearm, indigo jeans cuffed once, cream ribbed socks, short almond nails in milky nude, two thin silver rings on the RIGHT index and middle finger, dark brown centre parted shoulder length hair. Keep identical in every shot. Face visible only in SHOT 1 and SHOT 8.
- hand_behaviour: hands are already in contact whenever visible, fingers together, pads flat on leather, lid or floor
- camera_character: handheld phone throughout, real hand tremor, small live reframes

SEQUENCE LIST:
SHOT 1 (0 to 3s) HOOK. MCU, high angle as if arm held, frontal 0 degrees, 65 degree wide. Handheld with breathing. Girl on the bed edge x 25% to 80%, closed box at her chest, hands flat on the lid sides. Window light rakes the left of her face, catchlight in both eyes. She lifts the box toward the lens, then tips it out of frame bottom.
- dialog: Dialogue language: American English. Girl (light young alto, conversational) lifts the box a fraction, then says: 'Okay, these finally came.'
(MATCH CUT TO)

SHOT 2 (3 to 7s). CU, overhead 90 degrees. Handheld, drifting 5 cm closer. Hands lie flat on the lid, silver rings catching light. The lid lifts out of frame, white tissue folded over the pair beneath. Dry cardboard drag, one paper crackle. Fingertips settle on the fold.
(HARD CUT TO)

SHOT 3 (7 to 11s). MS, low angle at floor height, three quarter 45 degrees. Handheld, slow push toward the box. Fingers peel the tissue open and the rounded toe rises out of the paper, burnished edge catching light first. Tissue cracklre, then quiet. The hand closes on the shoe and lifts.
(CUT TO)

SHOT 4 (11 to 15s). CU, eye level, profile 90 degrees, 29 degree short telephoto, background compressed to a soft wash. Locked handheld while the shoe turns. One loafer held up, fingers round the heel counter, rotating 30 degrees toward the lens. The highlight travels the burnished edge from toe to heel, lug sole staying dark underneath. Rotation stops with the strap facing the lens.
(MATCH CUT TO)

SHOT 5 (15 to 19s) SIGNATURE. Macro, three quarter 45 degrees on the penny strap, close focus, depth of field a few centimetres. Handheld, tiny reframe, no travel. The strap fills frame. A fingertip already resting beside the keyhole draws along the strap edge. Perforated dots stay sharp, grain falls soft at frame edge. The finger lifts, frame settling on the keyhole.
(CUT TO)

SHOT 6 (19 to 22s). MS, low angle from the floor, three quarter 45 degrees on the lower leg. Handheld, static with breathing. Cream sock on the foot, jean cuff above the ankle. The foot slides in and the heel drops, lug sole taking weight on the oak, contact shadow forming under it. One dull sole tap. The foot presses down and holds.
(CUT TO)

SHOT 7 (22 to 26s). CU, POV straight down 90 degrees from standing height. Handheld, gentle body sway. Both loafers on the oak floor from above, cream socks and cuffed jeans above them, straps and keyholes reading clearly. A long highlight lies across both vamps. One foot pivots and settles.
(CUT TO)

SHOT 8 (26 to 30s) RESOLUTION. WS, very low angle with the lens 20 cm off the floor, frontal 0 degrees, 65 degree wide, floorboards running into the lens. Locked low handheld, framing identical at first and last frame. The floor sits empty, then Girl walks in from screen left, stops centre and turns to the lens. The loafers sit nearest camera and largest in frame, lug sole and burnished vamp both clear, a contact shadow travelling with each sole. Her head reads high and slightly softer than her feet. Two flat sole steps, then quiet. She settles her weight, one shoulder dropping.

TECHNICAL:
no subtitles, no captions, no on screen text, no background music, no extra dialogue beyond the line in SHOT 1

这是 Demo

当前是单页静态站:视频 + 指引 + 可复制提示词。后面可以做成多教程列表、账号、收藏提示词、一键复制、手机竖屏版等。

返回首页,看更多教程