APIMODELSAPIMODELS

Muse Video Cinematic & Story Prompts

More than 200 Muse Video prompts that creators posted on X after running them in Meta's Muse, each credited to its creator and linked to the original post, with the clip it produced where the post has one. Short prompts, and those whose authors invited reuse, are quoted in full; longer ones are excerpted, and the full text is one click away in the original post.

← All 204 Muse Video prompts

Last updated: · Muse Video API — pricing & playground · API docs

TL;DR

Muse Video is the video model in Meta's Muse AI family. Most prompts on this page were run inside Meta's Muse app, a chat agent that accepts very long prompts and plain instructions such as "turn the images in this post into a video". Through apimodels.app the same model is available as muse-video: text, one first frame or up to four reference images in; a 720p clip of 4 to 20 seconds out, 16:9 or 9:16; prompts up to 4,000 characters; $0.037 per requested second, failed tasks not charged, and not every request succeeds. The prompts are grouped by use: reference-sheet multi-shot cuts, boyfriend POV, selfie and vlog, dance, period pieces, fantasy and CG, long multi-clip scripts and one-liners.

Tears of joy against a plain background

Excerpt

柔和的正面漫射光均匀照亮人物面部,阴影浅淡过渡自然,背景受光均匀呈浅灰色. 温柔动容,带着喜极而泣的柔软酸涩感. 一位扎低马尾、穿白色细肩带背心的年轻女性在浅灰色纯色背景前,从平静注视逐渐展露笑意,最终眼含泪水双手合十,呈现喜极而泣的状态 室内纯色背景拍摄场景. 人物头肩部位,占据画面中心位置. 均匀的浅灰色纯色背景,无其他装饰或物体. 年轻东亚女性,黑色头发扎成低马尾,额前与脸侧垂有松散的湿感碎发,身着白色细肩带背心,面部皮肤带有自然粉晕,后期脸颊挂有透明泪水. Shot 1: 起始状态为人物正视镜头,表情平静柔和;过程中双眼向画面左侧转动,随后回视镜头,接着短暂闭合双眼;结束状态为睁开双眼,嘴唇开始向内抿起. 固定机位,静态拍摄,无镜头移动. close-up shot. …

This is an excerpt. Read the full prompt in the original post →
Full prompt:
blocks: Camera · 6 shots · 1166 characters in the original post
Watch on X
by @jarvis11xSource

A viral funny-animal scene, remade

Excerpt

生成视频:竖屏 9:16,时长约 10 秒,超写实真实拍摄质感,像用手机随手拍的家庭搞笑视频。固定机位,平视中景,镜头全程不动,无推拉摇移。 【场景】 温馨的中式现代家庭卧室。米白色墙壁,背景有一扇敞开的浅木色房门,门外是走廊。门边是一张原木色床头柜,上面有一盏暖白色布艺台灯(开着,暖光)和一瓶藤条香薰。床上是凌乱蓬松的白色被子和白色枕头,被面有自然的褶皱。窗外透进自然柔和的日光,室内光线明亮温暖,整体色调柔和生活化。 【主角】 一只小吉娃娃:浅棕红色短毛,大尖耳朵竖起,胸口有一小撮白毛,体型小巧。像人一样端坐在床上,双前爪捧着一部深色智能手机(黑色 iPhone 样式),专注地盯着屏幕。 一只金毛猎犬:体型比吉娃娃大很多,金黄色长毛,垂耳。身体放松地趴在床上,下巴搁在白色被子上,半眯着眼,但视线牢牢盯着手机屏幕,和吉娃娃一起看。两只狗的视线都落在同一部手机屏幕上。 …

This is an excerpt. Read the full prompt in the original post →
Full prompt:
sections: 场景 / 负面提示词 · 1441 characters in the original post
Watch on X
by @liangzhu_AISource

Osaka-style shopping street at dusk, tracking shot

黄昏时分的日本大阪风商业街,一位二十多岁的东亚女生走在人流中:深棕色长发扎成低马尾,穿米色风衣(腰间系带、后背有挡风片),脖子上围着米黑格纹拼棕色印花的围巾。镜头从她身后中景平稳跟拍,她匀速向前走,长发和风衣下摆随步伐轻轻摆动;街道左侧是电车轨道,一辆红白相间的电车缓缓驶过,右侧是鳞次栉比的小店:章鱼烧店、居酒屋挂着红色灯笼,红白配色的醒目招牌、SALE立牌、自动售货机、电线杆和纵横交错的电线,行人擦肩而过、各自忙碌不看镜头。走到一半,她缓缓回头看向镜头(先动肩膀再转头,马尾轻轻甩起),露出浅浅的微笑;镜头顺势缓缓推近成面部特写,她的笑容逐渐放大,最后开怀大笑,眼睛弯成月牙,几缕发丝被微风吹过脸颊。光线是黄昏暖金色,柔和侧逆光,轻微胶片颗粒,暖调干净通透,皮肤保留自然质感,背景人流虚化。一镜到底的云台稳定跟拍,只做一个连续运镜:跟拍+缓推。表情渐变自然:平静→浅笑→开怀大笑,动作链单一连贯。 负面要求:不要水印和文字叠加,不要突然转身或跳切,不要夸张美颜滤镜,背景行人不要盯镜头,主角不要换脸换装。

Watch on X
by @Super_NiusSource
Hover to preview · click for sound

Folk song in a small theatre concert

生成一段15秒、9:16竖屏的写实演唱会视频。 【人物与舞台】 成年女歌手小薇站在舞台中央,,手持黑色麦克风。小型剧场演唱会,暖金色聚光灯照亮人物,背景是雾蓝色灯光与缓慢流动的抽象水波。画面下方隐约可见观众剪影,整体自然、安静,有真实现场感。 【镜头】 一镜到底,摄影机平稳推进。 0—3秒:舞台正面中景,小薇随前奏轻轻摆动,准备开口。 3—10秒:缓慢推至半身景,捕捉自然演唱、呼吸和手部动作。 10—15秒:推进至面部近景,唱完最后一句后轻轻微笑,灯光和余音自然延续。 【演唱与声音】 温暖、略带颗粒感的女中音,中文咬字清晰,像讲故事一样演唱。木吉他搭配轻柔鼓刷,80 BPM,人声突出,保留适量现场混响。 0—1.5秒:吉他前奏。 1.5—7.5秒演唱:“泉水向南,慢慢地流淌。” 7.5—13.5秒演唱:“没见过的事,还想看一看。” 13.5—15秒:尾音渐落,伴随轻微观众掌声。 演唱口型与歌词同步,表情克制自然,人物面容、发型和服装全程一致。皮肤保留真实质感,画面干净,光线柔和。

by @RohoQinAISource

Early-morning café

生成一段 5 秒竖屏(9:16)电影感短视频,主题「清晨咖啡店」,全屏画面,不要胶片边框,不要 KODAK 等边框文字。 整体风格:超写实真人实拍,真实摄影、电影镜头拍摄感,非动漫、非 CG、非塑料 AI 感。清晨 7 点左右,低角度金色晨光,暖色调,Kodak Portra 400 胶片质感,轻微颗粒,真实皮肤纹理、真实玻璃反光、橱窗水汽。保留生活化不完美细节:街角落叶、人行道纹理、自行车、摩托车、纸箱、墙面涂鸦、报纸。 咖啡店规模:不要小铺面。要占多个开间的大店面街角咖啡店,木招牌、大落地窗,室内吧台、专业咖啡机、面包展示柜清晰可见,门口有外摆木桌椅。 镜头 1(0-1.7 秒):远景,晨光里的街角咖啡店,镜头缓慢推近。镜头 2(1.7-3.4 秒):特写慢动作,咖啡师拉花。动作与效果必须严格对应:奶缸嘴贴近杯面,奶流落点就是图案形成的位置;先左右小幅晃动,白色奶泡同步形成叶形纹理(rosetta),叶片由晃动逐层出现,不能凭空生成图案;最后奶缸抬高、细奶流从图案中心一笔收尾拉过,完成拉花。全程同一个杯子、同一个图案,连续形成。镜头 3(3.4-5 秒):中景,顾客双手捧着刚才那杯同款叶形拉花的咖啡,坐在窗边望向窗外晨光,轻轻微笑,画面柔和淡出。杯中图案与镜头 2 的成品保持一致。 转场:镜头之间柔和叠化。声音:轻爵士钢琴,配咖啡机的蒸汽声、杯碟轻响和清晨街道环境音。

Watch on X
by @0xPasswdSource

How to write great Muse Video prompts

  1. 1Lock the person with one reference sheet. The most-copied format here starts from a single composite image: a headshot on the left and a three-view outfit sheet on the right. The prompt points at it section by section (face from the headshot, outfit from the turnaround) and repeats that the face and outfit must not change in any shot. Through the API, send that sheet as a reference image, not as a first frame.
  2. 2Write the body in time steps and cut on the beat. Those prompts describe each shot in 0.3-second steps (weight shift, where each hand is, where the eyes go) and list the cut points, such as 1.80 s, 3.50 s and 5.10 s, against the beats of the music. "She poses" on its own gets you a stiff loop.
  3. 3For POV, make the camera a person. The boyfriend-POV series opens with the same rule in every episode: the camera is his eyes, and he never appears except for a hand or forearm. Then the action in 0–2 s / 2–5 s / 5–8 s / 8–10 s, one line of dialogue, and a handheld 9:16 phone look.
  4. 4Repeat a block for what must not change. The long English prompts open with a format lock (aspect ratio and source resolution) and a Master block that fixes the character and costume word for word and is pasted into every clip. Dance prompts add a tempo lock: the BPM and the timestamp of every downbeat.
  5. 5Ask for the flaws of a phone camera. Selfie and vlog prompts get their realism by asking for handheld shake, focus breathing, noise-reduction smear, platform compression and 720p, and by banning HDR, studio retouching and a face that stays sharp the whole time.
  6. 6Split anything longer than 20 seconds. Several prompts here describe films of 30 seconds to a minute. On the API each request is 4 to 20 seconds and up to 4,000 characters, so split by clip, keep the character block identical in each request, and edit the clips together.
  7. 7One-liners need the app, not the API. Instructions such as "turn the images in this post into a video" or "remake ten videos in the style of this clip" work in the Muse app because the agent opens the link and writes its own prompt. A video API only sees the text you send, so write the shots out.

Prompt formula

Format (aspect ratio, length) + who (locked by a reference image) + scene + light + camera position + shots written in time steps + style + sound + what to avoid

Example: 9:16, 10 seconds, handheld phone selfie. The woman from the reference image, same face, hair and outfit throughout. A quiet bookshop aisle in the late afternoon, warm window light from the left. 0–3 s: she pulls a book from the shelf and glances at the lens. 3–6 s: she lifts it to hide half her face and laughs. 6–10 s: she lowers it, looks straight into the lens and says "Found it." Slight handheld shake, natural skin texture, the soft sound of pages. Avoid: on-screen text, extra fingers, a second person.

Frequently asked questions

What is Muse Video?

Muse Video is the video model in Meta's Muse AI family. Most people use it inside Meta's Muse app, a chat agent: you describe the video, or paste a long prompt, and it produces the clip. Through apimodels.app the same model is available over an API as muse-video, with the limits listed below.

Who wrote these prompts, and why are some of them excerpts?

Each card credits the creator and links to the original post on X. A prompt is quoted in full when its author invited readers to use it, or when it is 600 characters or shorter; longer prompts are quoted from the beginning, word for word, and marked as excerpts, with the full text in the original post. If you wrote one of these prompts and want it removed or the credit corrected, email us or open an issue on the GitHub repository.

Where can I run these prompts?

In Meta's Muse app, where they were written, or through the API on apimodels.app as muse-video: text, one first frame or one to four reference images in; a 720p clip of 4 to 20 seconds out, 16:9 or 9:16. It is billed at $0.037 per requested second (4 seconds $0.148, 10 seconds $0.37, 20 seconds $0.74); failed tasks are not charged, and not every request succeeds. Clips of up to 10 seconds usually take 1 to 2 minutes.

Why does a long prompt from this page fail on the API?

The API takes prompts of up to 4,000 characters and clips of 4 to 20 seconds. Several prompts here are longer than that or describe a 30-second to one-minute film. Split them by clip, keep the character and costume block identical in every request, and edit the clips together.

How do I keep the same person across shots?

Send reference images: one to four photos of the person keep the face, hair and outfit without forcing the clip to start from any of them. A first frame instead makes the clip start from that exact picture. You can send one or the other, not both. The reference-sheet prompts on this page are written for a single composite image with a headshot and a three-view outfit sheet.

Do the one-line instructions work through the API?

No. Instructions such as "turn the images in this post into a video" rely on the Muse app opening the link and writing its own prompt. The API only receives the text and images you send, so write the shots out or use one of the full prompts instead.

Can I use these prompts commercially?

The prompts belong to their authors. Check the original post before reusing one commercially. For clips you generate yourself, check the terms of the service you generate them with.

Run these prompts on apimodels.app

Muse Video is live here alongside 150+ other image, video, audio and language models — one API, one key, pay as you go, with $0.10 of trial credit on sign-up.

This library is open source on GitHub

The whole collection is on GitHub as Markdown and JSON, one page per prompt, with credits and links to the original posts, plus a page to search and copy them. Prompts are quoted in full when the author invited reuse or the prompt is 600 characters or shorter; longer ones are excerpted from the beginning, with the full text in the original post. The videos stay with their creators and are not re-hosted. Our own notes, titles and data are licensed CC BY 4.0. Open an issue to correct a credit or have a prompt removed.

View on GitHub

Other prompt libraries