Sora 2 官方提示指南及案例 如何写出成功的视频提示词
官方视频提示指南。核心是提示词要写成一个完整的镜头描述——画面、物理、时间连贯性都交代清楚,模型才不会自由发挥。
- 先讲清模型能力边界:分辨率、时长、横竖屏,这些决定了你能要什么。
- 最大的变化是模拟真实物理,所以提示词里的动作描述要符合物理直觉。
- 提示词解剖那节把一条完整提示拆成结构,看每部分在干什么。
- Remix 功能允许在已有视频上精准改,不用从头再生成。
- 文末给中英文对照的模板合集,可以直接改着用。
OpenAI 宣布 Sora 2 与 Sora 2 Pro 正式开放至 API,为开发者带来新一代视频生成能力。两者分别支持 1280x720 与 1792x1024 的输出分辨率,可生成最长 12 秒的视频片段,支持横屏与竖屏模式。新增的 “Remix” 功能允许创作者对现有视频进行精准修改和重制。
Sora 2 的最大突破在于,它不再仅仅生成视觉画面,而是模拟真实物理世界的运作规律。模型能理解物体的运动、光线反射与声音传播,从而让生成的视频在物理层面上更真实、更自然。
这使得 AI 生成的画面几乎与真实拍摄难以区分,并赋予创作者对镜头、场景、光线和后期细节的高精度控制。
X2Twitter.com_VseaYuskgVn8M2NS_1024p.mp4
- Sora 2 与 Sora 2 Pro 正式加入 OpenAI API,支持生成 720p 与 1024p 视频。
- 新增 “Remix” 功能,可对现有视频进行精确编辑与改动。
- 每段视频最长 12 秒,支持横屏与竖屏模式。
- 当前版本 不支持真实人物视频或图像转视频。
- 图像:
- gpt-image-1-mini:比完整版便宜 80%,图像质量接近。
- gpt-realtime-mini:比完整版便宜 70%,用于实时语音交互。
Sora Video API 概览
Sora 的功能通过 OpenAI Video API 暴露给开发者, 该API包含五个主要端点(endpoints):

API 采用异步机制(asynchronous jobs),所有视频生成任务均在后台执行, 用户通过轮询或 webhook 监听生成进度。
模型层:Sora 2 与 Sora 2 Pro
OpenAI 提供两个版本:

- sora-2:主打 速度与可迭代性;适合快速创作和测试。
- sora-2-pro:主打 质量与稳定性;适合最终输出和商业制作。
Sora 2 官方提示指南及案例
这份官方指南教你:
如何像导演一样,用语言精准“指挥”AI摄像机。 如何通过清晰、具体、有层次的提示,让 Sora 生成接近你构想的视频。
Sora 提示的核心理念
把 Prompt 当作给摄影师或美术指导的拍摄说明书。
- 如果你写得太模糊,模型会“自由发挥”;
- 如果你写得太详细,模型会严格执行,但创意空间会变小;
- 关键是平衡:“明确意图 + 留一点想象空间”。
同时,Sora 每次生成都是“重新演绎”:
同样的提示,多次生成结果也会不同。 这不是Bug,而是创意的特性。
如何写出高质量 Prompt(Prompt Anatomy)
一个好的 Sora 提示词 = 一个小型分镜头脚本。 它应包含以下元素:

有效的提示词结构
清晰的提示词像是在把一镜画到分镜头板上。说明取景构图,标注景深,按“拍点”描述动作,设定灯光与色调。用少量独特细节固定住你的主体便于识别,而单一、可信的动作更易被跟随。
如果你要覆盖一个连续段落,也可以在单个提示词中描述多个镜头。这样做时,保持每个镜头块独立:一个机位设置、一个主体动作、一个光线配方。这样无论要生成短的独立片段,还是更长的连续时刻,你都更灵活。把每个镜头当作一个创意单元,你可以后期剪在一起,或者让它们在一次生成中连续播放。
- 简短提示词给模型更多创作自由。可以期待惊喜。
- 冗长、细节丰富的提示词会限制模型的自由度。它会努力遵循,但不一定总是完全可靠。
下面是一个简短提示词示例:
In a 90s documentary-style interview, an old Swedish man sits in a study and says, "I still remember when I was young."
这个提示词之所以常常有效,是因为:
- 90s documentary 设定了视频风格。模型会相应选择镜头、灯光和调色等变量。
- an old Swedish man sits in a study 用少量细节描述了主体和环境,让模型在人物与场景外观上有创作空间。
- and says, "I still remember when I was young." 描述了台词。Sora 往往能准确跟随。
这个提示词能稳定产出满足这些要求的视频。但它可能与你的具体设想不完全一致,因为许多细节仍未指明。例如,提示词没有描述时间、天气、服装、情绪、人物外观与年龄、机位、剪切、置景设计等。除非你写出这些细节,否则 Sora 会自行决定。
提示写作技巧
✅ Rule 1: Be specific, not vague
明确具体,不要模糊表达

✅ Rule 2: One clear action per shot
一个镜头一个动作
不要在一个Prompt里写太多镜头或复杂情节。 可以分多段生成后再剪辑。
不要只描述故事,而是描述镜头语言:
❌ Weak: The man walks, opens a door, sits, and talks.
✅ Strong: The man walks to the door and stops, hand reaching for the handle.
❌ 弱:「一个人在街上走」
✅ 强:「广角镜头,手持拍摄,一个人走过雨后的街道,路灯倒影在水中闪烁」✅ Rule 3: Control lighting & color
光线与色彩决定情绪

✅ Rule 4: Control motion & timing
控制动作节奏
描述节拍(beats)和时间感,帮助模型生成自然的运动:
❌ Actor walks across the room.
✅ Actor takes four steps to the window, pauses, and pulls the curtain in the final second.
演员走到窗边,停顿片刻,在最后一秒拉开窗帘。✅ Rule 5: “保持角色一致性”
描述人物时尽量保持相同措辞(性别、发型、衣着、气质), 否则模型可能生成不同角色。
✅ Rule 6: “短视频优于长视频”
Sora 在 4 秒片段中效果最佳。 建议:
拍多段 4 秒镜头,再在后期剪辑拼接。
走向“超细节”
对于复杂的电影化镜头,你可以超越标准结构,使用专业制作术语明确说明外观、机位设置、调色、声场乃至镜头意图。这类似导演给摄影组或特效组的简报。对镜头、滤镜、灯光、调色与运动的详细提示,有助于模型锁定非常具体的美学。
例如,你可以描述观众首先注意到什么、摄影平台与镜头、光线方向、色彩基调、材质与质感、画内声(diegetic)以及镜头时序。当你想匹配真实摄影风格(如 IMAX 航拍、35mm 手持、复古 16mm 纪录片)或在多镜头间保持严格连贯时,这种方法尤其有效。
示例
Format & Look
Duration 4s; 180° shutter; digital capture emulating 65 mm photochemical contrast; fine grain; subtle halation on speculars; no gate weave.
Lenses & Filtration
32 mm / 50 mm spherical primes; Black Pro-Mist 1/4; slight CPL rotation to manage glass reflections on train windows.
Grade / Palette
Highlights: clean morning sunlight with amber lift.
Mids: balanced neutrals with slight teal cast in shadows.
Blacks: soft, neutral with mild lift for haze retention.
Lighting & Atmosphere
Natural sunlight from camera left, low angle (07:30 AM).
Bounce: 4×4 ultrabounce silver from trackside.
Negative fill from opposite wall.
Practical: sodium platform lights on dim fade.
Atmos: gentle mist; train exhaust drift through light beam.
Location & Framing
Urban commuter platform, dawn.
Foreground: yellow safety line, coffee cup on bench.
Midground: waiting passengers silhouetted in haze.
Background: arriving train braking to a stop.
Avoid signage or corporate branding.
Wardrobe / Props / Extras
Main subject: mid-30s traveler, navy coat, backpack slung on one shoulder, holding phone loosely at side.
Extras: commuters in muted tones; one cyclist pushing bike.
Props: paper coffee cup, rolling luggage, LED departure board (generic destinations).
Sound
Diegetic only: faint rail screech, train brakes hiss, distant announcement muffled (-20 LUFS), low ambient hum.
Footsteps and paper rustle; no score or added foley.
Optimized Shot List (2 shots / 4 s total)
0.00–2.40 — “Arrival Drift” (32 mm, shoulder-mounted slow dolly left)
Camera slides past platform signage edge; shallow focus reveals traveler mid-frame looking down tracks. Morning light blooms across lens; train headlights flare softly through mist. Purpose: establish setting and tone, hint anticipation.
2.40–4.00 — “Turn and Pause” (50 mm, slow arc in)
Cut to tighter over-shoulder arc as train halts; traveler turns slightly toward camera, catching sunlight rim across cheek and phone screen reflection. Eyes flick up toward something unseen. Purpose: create human focal moment with minimal motion.
Camera Notes (Why It Reads)
Keep eyeline low and close to lens axis for intimacy.
Allow micro flares from train glass as aesthetic texture.
Preserve subtle handheld imperfection for realism.
Do not break silhouette clarity with overexposed flare; retain skin highlight roll-off.
Finishing
Fine-grain overlay with mild chroma noise for realism; restrained halation on practicals; warm-cool LUT for morning split tone.
Mix: prioritize train and ambient detail over footstep transients.
Poster frame: traveler mid-turn, golden rim light, arriving train soft-focus in background haze.
(译注:以上为专业格式示例,保持英文原样以便直接粘贴到工作流中。如需中文版本可另行说明。)
Prompt 模板合集(中英文对照)
🧱 基础模板(Descriptive Prompt Template)
[用日常语言写一段场景描述。描述角色、服装、场景、天气与其他细节。越贴合你的愿景越好。]
Cinematography:
Camera shot: [取景与角度,如 wide establishing shot, eye level]
Mood: [整体氛围,如 cinematic and tense / playful and suspenseful / luxurious anticipation]
Actions:
- [动作 1:清晰、具体的拍点或手势]
- [动作 2:在片段内的另一个明确拍点]
- [动作 3:再一个动作或一句对话]
Dialogue:
[若有对话,把简短自然的台词写在这里或放在 Actions 列表。保持简短以匹配片长。]示例 1
Style: 手绘 2D/3D 混合动画,柔和笔刷纹理,暖钨丝灯光,带触感的定格动画观感。整体美学呼应 2000 年代中期的绘本动画——温馨、不完美,充满机械质朴。淡淡水彩晕染与画布肌理;调色在暖与冷之间平衡;带电影感运动模糊以增强动画真实感。
在一间杂乱的工坊里,架子上堆满齿轮、螺栓和泛黄的蓝图。中央的木工作台上坐着一只小圆机器人,凹凸的机身用不匹配的金属板和旧漆修补过。它的大眼睛发着淡蓝光,一边紧张地摆弄一只嗡嗡作响的灯泡。空气里是轻微的机械嗡鸣,窗外雨点拍打,墙上的时钟稳稳走着。
Cinematography:
Camera: 中近景,缓慢推近;悬挂工具制造轻微视差
Lens: 35 mm 虚拟镜头;浅景深以柔化背景杂物
Lighting: 顶部道具灯暖色主光;窗外冷色溢光形成对比
Mood: 温柔、异想天开,带一点悬念
Actions:
- 机器人轻敲灯泡;滋滋火花闪现。
- 它一惊,手滑掉落灯泡,眼睛瞪大。
- 灯泡慢动作下坠;它在最后关头接住。
- 胸腔喷出一缕蒸汽——释然而自豪。
- 机器人小声说:“差点没了……不过我搞定了!”
Background Sound:
雨声、时钟滴答、轻柔机械嗡鸣、微弱灯泡嘶嘶声。
示例 2
Style: 1970 年代浪漫戏剧风,35 mm 胶片质感,保留自然炫光,柔焦与暖色晕影。轻微画幅摇曳(gate weave)与手持微抖营造复古亲密感。Kodak 风格调色;灯泡处轻微晕影;胶片颗粒与柔和暗角,提升年代感。
黄金时刻,砖砌公寓的屋顶化作一方小舞台。白床单在衣绳上随风摇摆,接住最后的阳光。几串不成对的彩灯头顶嗡嗡作响。一个穿红色真丝连衣裙的年轻女人赤脚起舞,卷发在余晖中泛着光。她的搭档——卷起衬衫袖子,吊带松垮——拍手打着节拍,笑容坦率。楼下,城市以车喇叭、地铁震颤与远处笑声低低地嗡鸣。
Cinematography:
Camera: 中广景,平视缓慢推进
Lens: 40 mm 球面镜头;浅景深将二人与天际线分离
Lighting: 金色自然主光配钨丝灯反射补光;彩灯形成边缘光
Mood: 怀旧、温柔、电影感
Actions:
- 她旋转,裙摆展开,接住一束阳光。
- Woman(轻笑):“看吧?今晚连城市都跟着我们跳。”
- 他上前接住她的手,把她一带,落入阴影。
- Man(含笑):“那是因为你领舞。”
- 床单掠过画面,短暂遮住天际线,又缓缓分开。
Background Sound:
只用自然环境声:轻风、布料拂动、街噪、远处含混的音乐。无额外配乐。详细案例
🎨 示例 1:纪录片访谈风格(Interview Style)
Prompt (EN+CN):
In a 90s documentary-style interview, an old Swedish man sits in a study and says, "I still remember when I was young."
在90年代纪录片风格的访谈中,一位年长的瑞典男子坐在书房里,说道:“我还记得年轻时的日子。”
Camera shot: medium close-up, static tripod.
镜头:中景特写,静态三脚架拍摄。
Lighting: warm desk lamp from right, soft daylight from window left.
光线:右侧台灯暖光,左侧窗户柔光。
Mood: nostalgic, intimate.
氛围:怀旧而亲密。
🎬 示例 2:电影级镜头(Cinematic Look)
Prompt (EN+CN):
Wide shot of a woman standing in heavy rain, holding a red umbrella under a streetlight. Raindrops shimmer as headlights pass behind her.
广角镜头:一名女子站在大雨中,手持红伞站在路灯下。远处车灯掠过,雨滴闪烁。
Camera: 35mm lens, low angle, slow dolly-in.
镜头:35mm,低角度,缓慢推进。
Lighting: single warm streetlight from above, cool reflection from wet asphalt.
光线:上方暖色路灯,地面湿反冷光。
Mood: cinematic, melancholic, rain-soaked.
氛围:电影感、忧郁、被雨淋湿的质感。
🌇 示例 3:自然风景(Landscape)
Prompt (EN+CN):
Aerial wide shot of mountains at sunrise, mist rolling over the valleys, golden light breaking through clouds.
航拍广角镜头:日出时的山峦,薄雾在山谷间流动,金色阳光穿透云层。
Camera: drone-like movement, slow pan from left to right.
镜头:无人机视角,从左到右慢速平移。
Lighting: natural sunlight with volumetric rays.
光线:自然阳光,带体积光。
Mood: majestic, serene, awe-inspiring.
氛围:宏伟、宁静、令人敬畏。
🧠 示例 4:角色动画(Character Animation)
Prompt (EN+CN):
A small robot sits in a cluttered workshop, tinkering with a glowing light bulb. Sparks fly as it adjusts the wires.
一个小机器人坐在杂乱的工坊里,摆弄着发光的灯泡。它调整电线时火花飞溅。
Camera: medium close-up, slow push-in.
镜头:中景特写,缓慢推进。
Lighting: warm key light from overhead, cool window spill.
光线:顶部暖光,窗外冷色补光。
Mood: gentle, curious, slightly tense.
氛围:温和、好奇、略带紧张。
Dialogue: Robot (softly): "Almost lost it… but I got it."
对白:机器人(轻声):“差点丢了……但我修好了。”
🪄 示例 5:复古电影风格(Vintage Film)
Prompt (EN+CN):
Style: 1970s romantic drama, 35mm film, warm halation and soft focus.
风格:1970年代浪漫电影,35mm胶片质感,暖色晕光与柔焦。
At golden hour, on a rooftop covered with laundry lines, a woman in a red silk dress dances barefoot while her partner claps along.
在金色时刻,一位穿红丝绸裙的女子在屋顶赤脚起舞,她的伴侣在旁边拍手微笑。
Lighting: golden sunlight with tungsten bounce.
光线:金色阳光配钨丝反光板。
Mood: nostalgic, joyful, cinematic.
氛围:怀旧、欢乐、电影质感。
使用图像参考(Use image references)
你可以在视频生成时提供一张 输入图像(input image) 来引导生成过程。 这张图像会被用作视频的第一帧(first frame)。
这种方式非常有用,特别是在以下场景中:
- 希望生成的视频保持品牌资产(brand asset)的一致视觉风格;
- 想要在视频中复现特定角色或人物形象;
- 需要在相同的环境或布景中延续画面。
在调用 POST /videos 接口时,将图像文件作为参数 input_reference 传入即可。 注意:
图像分辨率必须与目标视频的分辨率(size 参数)一致。
支持的输入文件格式包括:
- image/jpeg
- image/png
- image/webp
示例命令
以下是一个通过 cURL 调用 API 的示例, 展示如何提供图像作为参考输入:
curl -X POST "https://api.openai.com/v1/videos" \
-H "Authorization: Bearer $OPENAI_API_KEY" \
-H "Content-Type: multipart/form-data" \
-F prompt="她转过身,微笑着,然后缓缓走出画面。" \
-F model="sora-2-pro" \
-F size="1280x720" \
-F seconds="8" \
-F input_reference="@sample_720p.jpeg;type=image/jpeg"
Prompt: “She turns around and smiles, then slowly walks out of the frame.” 提示词:“她转身微笑,然后慢慢走出画面。” Prompt: “The fridge door opens. A cute, chubby purple monster comes out of it.” 提示词:“冰箱门打开了。一只可爱、胖乎乎的紫色怪物从里面走了出来。” .gif)
在这个例子中:
- prompt 定义了视频的描述性提示语(即剧情或镜头描述)。
这里的提示是:
> “她转过身,微笑着,然后缓缓走出画面。”
- model 指定使用的模型版本,这里使用高质量模型 sora-2-pro。
- size 设置视频分辨率(例如 1280×720)。
- seconds 设置视频时长(这里为 8 秒)。
- input_reference 提供视频起始帧图像文件。
🎬 实际用途:
通过这种方式,你可以:
- 确保视频在视觉上与品牌或角色形象保持一致;
- 生成“从静态图像开始”的视频场景;
- 用相同风格的图像批量生成系列视频(如广告、剧情片段等)。
对白与音效
- 对话需写在 Dialogue: 段落中;
- 每个角色的台词应清晰、简短;
- 4 秒视频建议只包含 1–2 句;
- 可以加入“背景声音”提示(如“雨声”“地铁隆隆”)。
示例:
A dim interrogation room.
A detective and suspect sit across the table.
Dialogue:
- Detective: "You’re lying. I can hear it in your silence."
- Suspect: "Or maybe I’m just tired of talking."背景声音描述示例:
The hum of espresso machines and the murmur of voices form the background.“Remix” 功能:像剪辑一样修改视频
如果你已经生成了一个视频,可以用一句话做局部修改:
“让背景光变成蓝色” “在后面加一只小狗跑过”
Sora 会复用原有画面结构,只更新你描述的部分。 这大幅减少了重新渲染的成本。
🧩 基本原理
在 Remix 请求中,你只需要:
- 提供一个已完成视频的 remix_video_id(即要修改的视频ID);
- 附带一个新的 prompt,描述希望进行的更改内容。
系统会:
- 复用原视频的视觉结构、连贯性与镜头布局;
- 只对提示中描述的部分进行修改;
- 在保持主体一致与镜头稳定的前提下,执行风格或场景变化。
✅ 最佳实践: 每次 Remix 仅进行一次明确的修改。 小范围、具体的调整(如颜色、光线、情绪)能最大限度地保留原画面质量, 并减少“AI伪影”(artifacts)或画面失真。
🧠 调用示例(使用 cURL)
curl -X POST "https://api.openai.com/v1/videos/<previous_video_id>/remix" \
-H "Authorization: Bearer $OPENAI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"prompt": "将画面的配色调整为青绿色、沙色和铁锈红,并加入温暖的逆光效果。"
}'参数说明:
- <previous_video_id>:要修改的已完成视频ID;
- prompt:用自然语言描述修改目标(如调整色调、添加光影、改变情绪氛围等);
- 系统会自动继承前一个视频的时长、分辨率和运动轨迹。
Original video Prompt: “Change the color of the monster to orange.” Prompt: “A second monster comes out right after.”
有效提示策略(Prompt Engineering)
OpenAI 建议的高质量视频提示结构:
“镜头类型 + 主体 + 动作 + 环境 + 光照”
示例:
- “Wide shot of a child flying a red kite in a grassy park, golden hour sunlight.”
- “Close-up of a steaming coffee cup on a wooden table, soft morning light through blinds.”
该层次化描述可显著提高输出一致性与物理连贯性。
示例演示
A high-speed cinematic race through a vast sci-fi landscape of twin suns and drifting asteroids. Sleek spacecrafts weave and twist between glowing debris fields and towering metallic canyons, their engines flaring with neon blue light. The camera tracks dynamically from cockpit to wide aerial shots, capturing motion blur, heat distortion, and lens flares as racers skim just meters above the surface. Sparks and vapor trails arc behind them as they bank through tight turns at breakneck speed. The soundscape roars with thrusters, wind, and faint radio chatter. In the distance, a colossal city hovers above the horizon, lights shimmering through cosmic dust. The scene is epic, cinematic, and kinetic—an adrenaline-fueled ballet of light, speed, and gravity-defying motion. Space-Race.mp4
Cinematic scene looking off the coast of Maui where a father and his daughter admire the ocean where they see whales breaching. The daughter is elated and can't contain her excitement. maui.mp4
A vast futuristic city curving upward into the sky, its buildings bending and connecting overhead in a continuous loop. Gravity shifts seamlessly along the curve, with sunlight streaming across inverted skyscrapers. The scene feels serene and awe-inspiring—earthlike fields and rivers running along the inner surface of a colossal rotating structure. Upside-Down-City.mp4
A direct continuation of the existing shot of a lone red fox trotting through a misty pine forest at dawn. Begin with the fox already mid-stride, camera tracking right at the same slow speed and eye-level angle. After three seconds, the fox pauses briefly, turns its head toward the camera with a quick ear flick, then resumes trotting forward. Maintain the 35 mm anamorphic lens with shallow depth of field, soft volumetric dawn light, faint morning fog, and subtle film grain. The forest should remain consistent in color grade and density, with the fox’s movement flowing naturally as if no cut occurred. fox-walk.mp4
Cinematic scene of a Zebra and her baby at a watering hole. The scene intensifies as a lion lurks in the tall grass. The mother and her baby make a run for it before the lion pounces. Zebra-Chase.mp4
A mycelium time-lapse under glass: lace-like threads map the soil, pulsing nutrients as glowing dots. <cinematography> Camera: macro cross-section, locked; subtle focus pulls Mood: scientific wonder, intimate, real </cinematography> <action> - Hyphae branch and fuse; glowing pulses race along new paths. - A tiny mushroom caps and unfurls in a breath. </action> <dialogue> Narrator (hushed): “The forest’s internet comes online.” </dialogue> mushroom-network.mp4
An enthusiastic 90s TV ad for Sora, the magical video creation tool. 90s-TV-Ad.mp4
A direct continuation of the existing shot of a chameleon crawling slowly along a mossy branch. Begin with the chameleon already mid-step, camera tracking right at the same close, eye-level angle. After three seconds, its eyes swivel independently, one pausing to glance toward the lens before it resumes moving forward. Maintain the 100 mm anamorphic lens with shallow depth of field, dappled rainforest light, faint humidity haze, and subtle film grain. The moss texture and background greenery should remain consistent, with the chameleon’s deliberate gait flowing naturally as if no cut occurred. chameleon.mp4
A cozy coffee shop interior, warm afternoon sunlight streaming through tall windows. Two friends (mid-30s, one wearing a denim jacket, the other in a mustard sweater) sit at a wooden table by the window, sharing a laugh. The camera is at medium shot, eye-level, slow push-in toward them. Shallow depth of field, soft golden lighting, slight film grain for a cinematic feel. cozy-coffee-shop-interior.mp4
A peaceful, cinematic scene of two sea otters floating side by side on calm, sunlit water. They hold hands together gently as they drift, their fur glistening with droplets that sparkle in the golden light. The camera lingers close, capturing soft ripples and the subtle rise and fall of their breathing. Reflections of the sky shimmer across the surface—hues of blue and amber blending like watercolor. The moment feels tender and intimate, a quiet portrait of love and trust in nature’s still embrace Sleeping-Otters.mp4
Indie café by a rainy window. Maya Kim (late-20s, Korean American, wavy dark-brown bob to jawline, light freckles) wears an emerald cardigan over a cream tee, thin gold necklace; chipped blue mug on table. She reads a paperback with a red spine by a hanging fern; neon “OPEN” sign reflection glows in the glass. Camera: medium, eye-level, slow dolly-in (0.25 m/s). She glances up toward the door (camera-left) and smiles. indie-cafe-rainy-window.mp4
官方文档与示例链接:https://platform.openai.com/docs/guides/video-generation#remix-completed-videos
官方提示词指南:https://cookbook.openai.com/examples/sora/sora2_prompting_guide
