Back to blog
Tutorials10 min read

Luxury Cars Wait. He Escapes in a Laundry Cart: An AI Drama Opening

On this page

Luxury cars line the prison entrance. Men in suits bow, waiting for someone important. The story sends him out hidden in a laundry cart instead. How do you turn that reversal into AI drama shots? This case uses actual references and generated clips from this case to explain asset selection, shot prompts and revision decisions, with a workflow you can apply in Panco.

22-second opening: spectacle and returned possessions

01Build an expectation before the cart breaks it

The opening needs to establish status. Cars, a formation and the gate build the spectacle; the property counter introduces a watch, a lighter and an old car key. These props carry story information as well as visual detail. The actual 22-second opening above was recorded as Seedance-2.5, all-purpose reference mode, 720p.

Decide what the audience should know at this point. Who is everyone waiting for? What do the returned possessions suggest about the man's past? Keep the public welcome and his private circumstances distinct so the cart can disrupt that expectation. Save hiding, leaving and meeting again for their own shot tasks.

02Give each of seven references a specific job

Organize characters, locations and props in your Panco project, then add only the images needed for the current video. This opening uses two locations, two named characters and three props. The gallery below makes those roles visible: identity, space and objects each need a concrete reference.

The source task uses image numbers assigned to uploaded references. Keep the original syntax when reading the source, then choose your own assets and check their numbers in your task. Reordering uploads means checking the text mapping again. A shot without a character does not need that character added just to fill a reference slot.

01 · Prison gate: cars, formation and architecture
01 · Prison gate: cars, formation and architecture
02 · Property counter: worktop and cool interior light
02 · Property counter: worktop and cool interior light
03 · Qin Zhan: identity and prison outfit
03 · Qin Zhan: identity and prison outfit
04 · Lighter: returned-property close-up
04 · Lighter: returned-property close-up
05 · Car key: an object tied to the past
05 · Car key: an object tied to the past
06 · Gold watch: focus on the counter
06 · Gold watch: focus on the counter
07 · Da Shan: the greeter outside
07 · Da Shan: the greeter outside
Seven references and their shot roles
ReferenceRole in the frame
1Prison gate: cars, formation and architecture
2Property counter: worktop and cool interior light
3Qin Zhan: identity and prison outfit
4Lighter: returned-property close-up
5Car key: an object tied to the past
6Gold watch: focus on the counter
7Da Shan: the greeter outside

03Write Seedance prompts around visible shot tasks

The original prompt repeats many style adjectives, but its actionable instructions are the reference mapping, location, blocking, shots, actions and dialogue. Expand it below and compare it with the clip. Associate your own images before using the text in a new task.

Expand the original generation prompt
Expand the original generation promptOriginal Chinese input
国产短剧风格,超写实风格,极致的场景细节以及人物面部特写,情绪大开大合,保持角色一致性。真人实拍电影质感,无任何Al生成痕迹,运镜丝滑,电影级表演,人物不要僵硬无表情。全程表演丝滑、不僵硬、不卡顿、表情细节表演真实到位。空间关系排布正确。整体采用定焦镜头,环境光影色调统一,保持环境信息不变,脸部细节清晰,采用真人实拍质,感发丝根根分明,表情灵动自然,面容细腻渲染。头发跟着人物动作微微晃动,呈现电影级写实风格,禁用卡通元素与漫画线条,注重真实光影与质感表现,整体风格生活化、自然且不夸张,人物说完台词自然停顿留白,画面风格为原生无字幕短视频,无信息条,无字幕条。保持无字幕,避免画面生成字幕。所有对白和音效仅通过声音呈现,画面中不叠加任何文字或字幕,禁止生成背景音乐,可以出现恰当适宜的音效。
皮肤:必须清晰可见毛孔、微细血管、自然红润肤色和微小皱纹。严禁出现任何塑料感或陶瓷般光滑肌。动作:模拟肢体重量感,走路动作必须体现双脚落地的力度传导。服装(衣角)需随身体惯性自然摆动,拒绝僵硬和漂浮。
眼神:必须包含清晰的眼神光(catchlight)和自然的聚焦(focus),面部需带有微妙的情绪微表情,眼神灵动,拒绝空洞。(严格规避 AI 表情僵硬、动作机械化通病,所有反应基于真实人体生理逻辑,微表情与肢体细节颗粒度拉满,氛围感与代入感拉满)(注意不要出现两个相似的人物)

@图片1
 是监狱大门口。
@图片2
 是监狱内务窗口。
@图片7
 是大山哥。
@图片3
 是秦战。
@图片6
 是劳力士。
@图片5
 是奔驰车钥匙。
@图片4
 是打火机。


场景:白天,阳光直射的监狱大门口
@图片1
 。
人物:大山哥、宏盛集团员工数十、
人物站位:豪车沿街道排开,西装汉子整齐列队于大门两侧,


镜号1:【航拍超广角俯瞰】阳光直射,高墙电网的监狱大门前景压迫式占据画面下部,一列黑色豪车沿街道两侧排开半条街,数十名黑西装汉子整齐列队如黑色铁壁,画面顶部远景可见监狱高塔,顶光硬射,地面投下锐利阴影;
镜号2:【低角度仰拍特写】西装汉子们黑色皮鞋整齐并拢,西裤笔挺,双手背在身后,阳光在地面上投下刀刻般的阴影线,氛围肃杀压抑,硬光顶照;
镜号3:【中景平视缓推】监狱大门缓缓向两侧滑开,金属摩擦声沉闷低回,
镜号4:【过肩镜头转全景】从队列后方过肩望去,大门完全打开,三十名西装汉子整齐划一地缓缓低头鞠躬,动作如出一辙,




场景:白天,日光灯管惨白照明的监狱内务窗口
@图片2
 。
人物:狱警、秦战
人物站位:秦战背对镜头站着,狱警坐在桌后


镜号1:【中景平视】监狱内务窗口,日光灯管投下冷光,台面空旷,狱警坐在桌后整理三样物品,
@图片3
 秦战背对镜头站在狱警面前,
镜号2:【近景特写跟焦】狱警将
@图片6
 纯金劳力士手表放在铁灰色台面上,狱警开口:"现在归还你入狱时的私人物品,";
镜号3:【微距特写】
@图片4
 打火机被推到台面上,狱警继续:"纯金劳力士手表一支,镶钻的zippo打火机";
镜号4:【俯拍特写】一把泛着旧时代光泽的奔驰车钥匙
@图片5
 被放在台面上,金属表面有细微磨损痕迹,狱警说:"二十年前的奔驰车钥匙,";
镜号5:【特写】
@图片3
 秦战一只手接过物品,手指骨节分明,掌心握住三样物品。
镜号6:【过肩特写】
@图片3
 秦战只露背影,手握物品收进口袋,肩膀微微一动,狱警耸肩补充,语气带几分同情:"但车已经被你老婆卖了……";

At the property counter, the room reference defines the space, Qin Zhan defines identity, and the three object images distinguish the watch, lighter and key. Presenting one object with its corresponding line makes the information readable. Check object shapes, hand contact and whether spoken information matches the visible object.

For a new prompt, start with an observable action, then add the camera: the officer places the watch on the counter; Qin Zhan reaches for it. That gives a clearer shot task than repeated requests for cinematic quality. Keep location and light consistent across cuts, changing shot size where the story needs it.

04From 25 seconds to 9: narrow the story task

Two real tasks cover this passage. The older version runs 25 seconds with seven references and includes the gate, street corner, departure and Qin Zhan appearing. The newer version runs 9 seconds with five references and concentrates on the cart passing the welcome party. Play both below.

Duration, references and story scope all changed. The useful lesson is how the task was narrowed. The newer result shows the cart, a character close-up, the formation and the gate; check whether those images communicate the intended reversal. The longer version also contains street and reveal material that can be considered separately.

When a result does not tell the story clearly, choose the one action this generation must deliver. Assign hiding, passing the crowd, reactions and the street-corner reveal to separate shots. Keep usable intervals and regenerate the missing connection.

Older task: 25 seconds, seven references
Newer task: 9 seconds, five references

Older task: 25 seconds, seven references

Seedance-2.5 · 720p

Expand excerpts from the generation prompt
Expand excerpts from the generation promptOriginal Chinese input
国产短剧风格,超写实风格,极致的场景细节以及人物面部特写,情绪大开大合,保持角色一致性。真人实拍电影质感,无任何Al生成痕迹,运镜丝滑,电影级表演,人物不要僵硬无表情。全程表演丝滑、不僵硬、不卡顿、表情细节表演真实到位。空间关系排布正确。整体采用定焦镜头,环境光影色调统一,保持环境信息不变,脸部细节清晰,采用真人实拍质,感发丝根根分明,表情灵动自然,面容细腻渲染。头发跟着人物动作微微晃动,呈现电影级写实风格,禁用卡通元素与漫画线条,注重真实光影与质感表现,整体风格生活化、自然且不夸张,人物说完台词自然停顿留白,画面风格为原生无字幕短视频,无信息条,无字幕条。保持无字幕,避免画面生成字幕。所有对白和音效仅通过声音呈现,画面中不叠加任何文字或字幕,禁止生成背景音乐,可以出现恰当适宜的音效。
皮肤:必须清晰可见毛孔、微细血管、自然红润肤色和微小皱纹。严禁出现任何塑料感或陶瓷般光滑肌。动作:模拟肢体重量感,走路动作必须体现双脚落地的力度传导。服装(衣角)需随身体惯性自然摆动,拒绝僵硬和漂浮。
眼神:必须包含清晰的眼神光(catchlight)和自然的聚焦(focus),面部需带有微妙的情绪微表情,眼神灵动,拒绝空洞。(严格规避 AI 表情僵硬、动作机械化通病,所有反应基于真实人体生理逻辑,微表情与肢体细节颗粒度拉满,氛围感与代入感拉满)(注意不要出现两个相似的人物)

@图片1
 是监狱大门口。
 
@图片7
 是街道口
@图片2
 是大山哥。
@图片3
  是秦战。
 
@图片4
 是老头。
@图片5
 是老头推着的
@图片3
 躲藏在里面的洗衣筐。




场景:白天,
@图片1
 阳光直射的监狱大门口。
人物:
@图片2
 大山哥、宏盛集团员工数十、
@图片4
 老头
人物站位:众人列队于大门两侧鞠躬,老头推洗衣筐从门内走出


镜号1:【全景仰拍】
@图片6
 画面闪白切至监狱大门外景,阳光刺眼,大门完全打开,数十名西装汉子齐声高呼,声浪在空旷的门前广场回荡:"东哥";
镜号2:【中景跟拍】一个佝偻老头
@图片4
 推着一辆洗衣筐
@图片5
 缓缓从大门内走出,筐上盖着旧衣服,与众人肃穆列队的场面形成荒诞反差,侧逆光勾勒老头轮廓;
镜号3:【近景特写快切】
@图片2
 的表情瞬间呆滞,嘴微张,眼神迷茫,旁边的人同样面面相觑,暖色阳光打在他们错愕的脸上;
镜号4:【全景平视】众人从呆愣中恢复原样,重新挺直站好,目光仍追随着老头背影,气氛尴尬沉默,顶光直射;
镜号5:【远景跟拍渐远】
@图片4
 老头推着
@图片5
 洗衣筐缓缓沿街道离去,背影在刺目阳光中渐行渐远,


场景:白天,
@图片7
 街道口。
人物:
@图片3
 秦战、
@图片4
 老头
人物站位:老头推洗衣筐站在街角,秦战藏在洗衣筐内


镜号1:【中景平视】街道口转角处,
@图片4
 老头推着
@图片5
 洗衣筐停下,左右张望确认四周无人,
@图片4
 老头弯腰掀开筐上的衣服;
镜号3:【近景仰拍】洗衣筐内,
@图片3
 秦战缓缓钻出,双手撑筐沿起身,侧逆光勾勒他的轮廓,秦战:"谢谢。";
镜号4:【中景侧拍】
@图片4
 老头一脸困惑地打量秦战,手指着洗衣筐:"你说你这身份,钻这里面干嘛啊?";
镜号6:【特写】
@图片4
 老头愣住,嘴微张,欲言又止,

Newer task: 9 seconds, five references

Seedance-2.5 · 720p

Expand the original generation prompt
Expand the original generation promptOriginal Chinese input
国产短剧风格,超写实风格,极致的场景细节以及人物面部特写,情绪大开大合,保持角色一致性。真人实拍电影质感,无任何Al生成痕迹,运镜丝滑,电影级表演,人物不要僵硬无表情。全程表演丝滑、不僵硬、不卡顿、表情细节表演真实到位。空间关系排布正确。整体采用定焦镜头,环境光影色调统一,保持环境信息不变,脸部细节清晰,采用真人实拍质,感发丝根根分明,表情灵动自然,面容细腻渲染。头发跟着人物动作微微晃动,呈现电影级写实风格,禁用卡通元素与漫画线条,注重真实光影与质感表现,整体风格生活化、自然且不夸张,人物说完台词自然停顿留白,画面风格为原生无字幕短视频,无信息条,无字幕条。保持无字幕,避免画面生成字幕。所有对白和音效仅通过声音呈现,画面中不叠加任何文字或字幕,禁止生成背景音乐,可以出现恰当适宜的音效。
皮肤:必须清晰可见毛孔、微细血管、自然红润肤色和微小皱纹。严禁出现任何塑料感或陶瓷般光滑肌。动作:模拟肢体重量感,走路动作必须体现双脚落地的力度传导。服装(衣角)需随身体惯性自然摆动,拒绝僵硬和漂浮。
眼神:必须包含清晰的眼神光(catchlight)和自然的聚焦(focus),面部需带有微妙的情绪微表情,眼神灵动,拒绝空洞。(严格规避 AI 表情僵硬、动作机械化通病,所有反应基于真实人体生理逻辑,微表情与肢体细节颗粒度拉满,氛围感与代入感拉满)(注意不要出现两个相似的人物)

@图片1
 是监狱大门口。
  是街道口
@图片2
 是大山哥。
  是秦战。
 
@图片3
 是老头。
@图片4
 是老头推着的 躲藏在里面的洗衣筐。




场景:白天,
@图片1
 阳光直射的监狱大门口。
人物:
@图片2
 大山哥、宏盛集团员工数十、
@图片3
 老头
人物站位:众人列队于大门两侧鞠躬,老头推洗衣筐从门内走出




镜号2:【中景跟拍】一个佝偻老头
@图片3
 推着一辆洗衣筐
@图片4
 缓缓从
@图片5
 大门内走出,筐上盖着旧衣服,与众人肃穆列队的场面形成荒诞反差,侧逆光勾勒老头轮廓;
镜号3:【近景特写快切】
@图片2
 的表情瞬间呆滞,嘴微张,眼神迷茫,旁边的人同样面面相觑,暖色阳光打在他们错愕的脸上;
镜号4:【全景平视】众人从呆愣中恢复原样,重新挺直站好,目光仍追随着老头背影,气氛尴尬沉默,顶光直射;
镜号5:【远景跟拍渐远】
@图片3
 老头推着
@图片4
 洗衣筐缓缓沿街道离去,背影在刺目阳光中渐行渐远,

05Asked for backs, got a face: review the pickup

The latest 4-second task requested the two characters' backs at a street corner. The actual result moves from a frontal view toward Qin Zhan's face. It is a candidate for a reveal or reaction, but it does not deliver the planned rear-view ending.

Expand the original generation prompt
Expand the original generation promptOriginal Chinese input
国产短剧风格,超写实风格,极致的场景细节以及人物面部特写,情绪大开大合,保持角色一致性。真人实拍电影质感,无任何Al生成痕迹,运镜丝滑,电影级表演,人物不要僵硬无表情。全程表演丝滑、不僵硬、不卡顿、表情细节表演真实到位。空间关系排布正确。整体采用定焦镜头,环境光影色调统一,保持环境信息不变,脸部细节清晰,采用真人实拍质,感发丝根根分明,表情灵动自然,面容细腻渲染。头发跟着人物动作微微晃动,呈现电影级写实风格,禁用卡通元素与漫画线条,注重真实光影与质感表现,整体风格生活化、自然且不夸张,人物说完台词自然停顿留白,画面风格为原生无字幕短视频,无信息条,无字幕条。保持无字幕,避免画面生成字幕。所有对白和音效仅通过声音呈现,画面中不叠加任何文字或字幕,禁止生成背景音乐,可以出现恰当适宜的音效。
皮肤:必须清晰可见毛孔、微细血管、自然红润肤色和微小皱纹。严禁出现任何塑料感或陶瓷般光滑肌。动作:模拟肢体重量感,走路动作必须体现双脚落地的力度传导。服装(衣角)需随身体惯性自然摆动,拒绝僵硬和漂浮。
眼神:必须包含清晰的眼神光(catchlight)和自然的聚焦(focus),面部需带有微妙的情绪微表情,眼神灵动,拒绝空洞。(严格规避 AI 表情僵硬、动作机械化通病,所有反应基于真实人体生理逻辑,微表情与肢体细节颗粒度拉满,氛围感与代入感拉满)(注意不要出现两个相似的人物)
 
 
@图片4
 是街道口
 
@图片1
  是秦战。
 
@图片2
 是老头。
@图片3
 是老头推着的洗衣筐。




场景:白天,
@图片4
 街道口。
人物:
@图片1
 秦战、
@图片2
 老头
人物站位:秦战站在街角望向远方,老头站在秦战身侧






镜号2:【空镜】落日余晖 ,拍摄
@图片1
 
@图片2
 二人的背影

For an ending built around departure, write a dedicated rear-view task. The revised prompt below uses daylight consistently, places the camera behind both characters and keeps them from turning. It also removes the conflict between an empty shot and two people in frame. This is a plan for the next generation; expand the original input to compare it with the existing clip.

Actual 4-second result: a frontal push-in

Revised prompt for the next pickup

Revised prompt for the next pickupEnglish prompt
Reference mapping: @图片1 = Qin Zhan; @图片2 = the older man; @图片3 = the laundry cart; @图片4 = the street corner.
Daylight at the street corner in @图片4. Keep the same clothing and location details as the selected references.
One continuous 4-second rear-view shot. The camera stays behind @图片1 and @图片2 at shoulder height. At the start, both stand beside @图片3, facing down the street. They slowly walk away, keeping their backs to the camera. The older man pushes the cart. The camera remains stationary.
End with both figures farther down the street, still seen from behind. Neither turns toward the camera. No cuts to faces, no frontal views, no sunset lighting, no extra people, no on-screen text or background music. Quiet street ambience only.

06Organize the next production pass in Panco

In Panco, separate the story into the welcome, the unexpected cart and the street-corner reveal. Prepare each passage's characters, locations and props. Keep names consistent and save costume versions separately. Describe subject, action and starting and ending positions in the storyboard, then attach the references that shot needs.

Arrange candidate clips on the infinite canvas with notes on usable intervals and missing shots. In editing, establish the order of revelations before adjusting dialogue, ambience and pauses. A departure shot must begin where the previous shot ends; matching character names alone does not establish continuity.

An AI drama opening becomes easier to revise when assets are reusable, each generation has one clear task and the footage guides your choices. Continue with the full Panco workflow for entry points, or the character and prompt guides for specific production problems.

Have a Story Idea? Make Your First AI Short Film with PancoAI Character Consistency: Fix Changing Faces and Clothes in PancoSeedance Prompts: Fix Unclear Actions, Camera Moves and Dialogue

07AI drama production questions

Do more references make characters more consistent?

Each image needs a clear role. Conflicting looks of the same person make the instruction ambiguous. The seven opening references cover identity, locations and props; select a fresh set for each shot's needs.

How do I reference characters and images in Panco?

Select actual assets in the relevant editor first. Named storyboard references and numbered images in a video task use different forms. This article preserves the source task's image numbers. Recheck the mapping whenever images are replaced or reordered.

Should I keep or regenerate a shot?

Check whether it delivers its story task, then inspect characters, actions, props and continuity. The 4-second result offers a frontal close-up rather than the requested rear view. Consider it for a reveal; regenerate that specific shot if the ending needs backs walking away.

Your next story starts here.

Turn your inspiration into your first shot.

Start creating
All articles