跳到主要内容
AI 视频模型
Runway

Runway

用文字描述生成短视频,或让单张图片成为起始素材。先选有效的时长与分辨率组合,再按任务目标编写画面和运动描述。

输入方式
2 种:文字或单张图片
生成预扣
依组合为 12 或 30 积分
提示词上限
1,800 字符
图片要求
最多 1 张 JPG/JPEG/PNG, 10 MiB 以内
快速了解
  • 从文字或单张图片开始,图生同样需要清楚的提示词。
  • 有效组合与预扣额逐项对应,1080p 仅支持 5 秒。
  • 图片仅限 JPG/JPEG/PNG,需完成上传及元数据校验。
  • 符合完整资格的成功 720p 结果,可另开 Extend 续写任务。

相关查询: Runway 怎么用 · Runway 文生视频 · Runway 图生视频

示例

Cinematic scene

STYLE: Miniature Tiny-Person FX, photorealistic live-action, cinematic macro photography, ultra-shallow depth of field, bright modern convenience store, realistic fluorescent lighting, ultra-detailed 8K, vertical 9:16, dynamic viral short-video style. DURATION: 15 seconds. SETTING: A huge modern convenience store packed with towering shelves of colorful snack bags, instant noodles, bottled drinks, soda cans, candy, cereal boxes, refrigerators, shopping baskets, and hanging price tags. Everything feels enormous from the perspective of a palm-sized woman. CHARACTER: One athletic miniature woman (1:15 scale), identical appearance throughout: lightweight adventure jacket, cargo pants, sneakers, utility backpack, same face, hairstyle, proportions, and size in every shot. 00:00–00:03 — Snack Shelf Swing Standing on a giant chip bag, she sprints and leaps toward a hanging price tag, grabs it with both hands, swings across the aisle, and launches onto the next shelf. The tag swings behind her. Camera: Low-angle macro tracking. SFX: Plastic crinkles, tiny footsteps, rushing air, swinging tag. 00:03–00:06 — Bottle Cap Parkour She lands on a giant drink bottle and parkours across colorful bottle caps. One bottle wobbles beneath her; she regains balance and performs a forward flip onto the next shelf. Camera: Fast side-tracking macro. SFX: Footsteps, bottle creaks, liquid slosh, hard landings. 00:06–00:09 — Shopping Basket Zipline She grabs a hanging promotional cord and ziplines above snacks and noodles. Releasing at the perfect moment, she lands inside a moving shopping basket among giant chocolate bars and noodle cups. Camera: Overhead to low-angle transition. SFX: Zipline whistle, wind, basket wheels, impact. 00:09–00:12 — Runaway Soda Can A giant soda can rolls inside the basket. She jumps onto it, balances like a surfer, and rides it across the polished floor, narrowly avoiding shoppers' shoes, baskets, and scattered snacks. Camera: Ultra-low floor-level macro tracking. SFX: Rolling metal, wind, store ambience. 00:12–00:15 — Victory Finale Using the rolling can as a launch ramp, she leaps to a freezer handle, swings up, and stands triumphantly on top of the freezer. The massive store stretches behind her as sparkling particles appear and large golden text reads: "CHALLENGE COMPLETE" She raises one fist while the camera slowly pulls back, emphasizing her tiny scale. Camera: Cinematic pullback. SFX: Leap whoosh, metallic grab, victory chime, sparkles, store ambience. VISUAL REQUIREMENTS: Maintain a consistent 1:15 miniature scale, realistic physics, smooth continuous action, photorealistic packaging, authentic reflections, cinematic macro DOF, and consistent character identity. No animation, cartoons, narration, subtitles, outfit changes, or scale inconsistencies.

Cinematic scene

Create an exceptionally realistic 30-second live-action cinematic short film in 16:9. It must look entirely filmed in the real world with real actors, practical locations and natural camera operation. Nothing should resemble AI-generated imagery or a polished advertisement. MAIN CHARACTER One consistent 25-year-old man, slightly untidy black hair, faint natural stubble and a slim average build. He wears a faded charcoal T-shirt, dark olive overshirt, loose black trousers, worn off-white sneakers and a black canvas shoulder bag. Preserve his exact face, hairstyle, age, body, outfit and proportions throughout. RECURRING OBJECT One small square pale-blue pastry box tied with thin red cotton string. Preserve its exact size, colour and appearance in every shot. The box contains one simple vanilla cupcake with white icing and one thin unlit birthday candle. Do not reveal the contents before the specified moment. 0–3.5 SECONDS Begin in a modest, genuinely lived-in city apartment during overcast late afternoon. A friend appears to be filming from knee height on a handheld cinema camera. The man sits on the edge of his bed tying one sneaker. He briefly looks toward the camera and says naturally, with quiet humour: “Day ten thousand of living life. I thought it would feel bigger.” Perfect natural English speech, subtle facial movement and accurate lip synchronization. No exaggerated acting. 3.5–7 SECONDS The camera follows him through the apartment in one loose handheld movement. He picks up the pale-blue pastry box and his canvas bag from the kitchen counter. Include ordinary background imperfections: an unwashed coffee cup, slightly crooked photographs, cables near the desk and soft window light. He leaves the apartment. The dark apartment door briefly fills the frame, creating a clean practical transition. 7–11 SECONDS Cut as a similarly dark elevator door opens. He walks out into a real apartment lobby and continues toward the street. The camera follows several steps behind him. Maintain believable autofocus corrections, footsteps, clothing movement and environmental sound. 11–15 SECONDS Long-lens side view from across a wet city street. He crosses a busy pedestrian crossing carrying the same box. Light drizzle, vehicles waiting naturally, umbrellas, imperfect pedestrian spacing and realistic reflections on the road. Nobody looks at the camera. Do not isolate him with artificial depth of field. 15–19 SECONDS A passing city bus briefly obscures the entire frame. When it clears, reveal the same man arriving outside a small neighbourhood laundromat during early evening. This is a motivated hidden cut, not a morph. The camera follows him physically through the open laundromat door. 19–23 SECONDS Wide, quiet composition inside the laundromat. Slightly harsh fluorescent ceiling lights, rain-marked front windows, mismatched plastic chairs and several washing machines rotating at different speeds. He places a small load of clothes into one machine, sits down and rests the closed pale-blue box on his lap. His behaviour remains casual and understated. 23–27 SECONDS He unties the red string and opens the box for the first time, revealing one cupcake with one unlit candle. An elderly female laundromat attendant in her early 60s, with grey hair in a low bun and a burgundy cardigan, notices from nearby. Without speaking, she approaches, strikes a real wooden match and gently lights the candle. Show the match resisting for a fraction of a second before igniting. Use a close natural shot of her hand, the flame and his surprised half-smile. Anatomically correct hands and realistic fire. 27–30 SECONDS He watches the tiny candle with a restrained, slightly emotional smile. The camera slowly dollies backwards through the already-open laundromat doorway and onto the wet pavement, leaving him and the attendant framed naturally inside the warm laundromat. The city remains cool and rainy outside. Over the final pullback, his quiet voice says: “Maybe the ordinary days were the big thing.” Finish on the candlelight reflected faintly in the rain-streaked laundromat window. Hold the final image steady for the last half-second. VISUAL STYLE Authentic independent short film, natural 24fps motion, realistic shutter blur, restrained cinematic colour, practical lighting, subtle sensor grain, occasional focus breathing, imperfect handheld movement and detailed environmental sound. Use physically possible camera movements and clean hard cuts or motivated foreground wipes. The character must behave as though the camera is a familiar friend, not a commercial crew. AUDIO Begin with quiet apartment ambience and natural dialogue. Gradually introduce an understated warm instrumental track with soft piano, brushed percussion and low analogue texture. Preserve footsteps, elevator movement, rain, traffic, laundromat machines, the match strike and the small flame. Music must never overpower the spoken lines. EXCLUDE No subtitles, captions, logos or on-screen text. No beauty filtering, plastic skin, excessive orange-and-teal grading, surreal lighting, slow motion, speed ramps, artificial bokeh, floating objects, morphing locations, duplicated people, changing clothes, changing pastry box, extra fingers, malformed hands, random staring pedestrians, exaggerated smiles, melodramatic acting or glossy advertising imagery.

Fashion portrait

Cinematic scene

Create a 30-second ultra-cinematic romantic short film featuring the same two clearly adult characters, both around 24–26 years old, throughout the entire video. The woman is elegant, expressive, and naturally beautiful, wearing a long dark coat over a soft cream outfit, with loose natural hair. The man is stylish and understated, wearing a charcoal jacket, dark trousers, and a simple watch. Maintain their exact same faces, hairstyles, outfits, body proportions, and identity in every shot. VISUAL CONCEPT Two strangers keep passing through the same city without noticing each other. They appear in the same places only seconds apart, creating the feeling that fate keeps bringing them close. The entire film builds toward one final moment when they finally see each other. 0–3 SEC — VIRAL HOOK Nighttime city street after light rain. Extreme cinematic close-up of the woman walking past camera. At the exact same moment, the man walks in the opposite direction behind her. They pass back-to-back only inches apart without seeing each other. Camera performs a smooth 180-degree rotation between them. For one second, both turn their heads slightly… …but just miss each other. Rain droplets glow under streetlights. --- 3–7 SEC — TRAIN WINDOW The woman sits inside a moving city train beside the window. Outside, the man stands on the opposite platform. Camera focuses through the glass reflection. For a brief moment, their faces visually overlap in the train window reflection. The train starts moving before either notices. Slow cinematic push-in. --- 7–11 SEC — COFFEE SHOP NEAR MISS Warm stylish café. The woman leaves through the front door carrying a takeaway coffee. Half a second later, the man enters through the same door. As they pass, the edge of her coat lightly moves from the air of the opening door. He briefly looks back. She has already disappeared into the crowd. -- 11–15 SEC — STREET CROSSING Golden evening light. Wide cinematic city intersection. The woman crosses from left to right. The man crosses from right to left on the opposite side. A yellow taxi passes between them and temporarily blocks the view. When the taxi clears… both have already passed. Use a beautiful overhead shot showing how close their paths were. --- 15–19 SEC — RED SCARF MOMENT Wind suddenly picks up. A small deep-red scarf tied to the woman's handbag becomes loose and flies away. Slow-motion close-up as the scarf moves through the air. The man reaches out and catches it naturally with one hand. He looks toward the direction it came from. --- 19–23 SEC — THE CHASE Music begins building emotionally. The man walks quickly through the crowd holding the red scarf. Fast cinematic cuts: His shoes moving across wet pavement. Her silhouette ahead. Streetlights reflecting in puddles. His hand holding the scarf. Her hair moving in the evening wind. Camera movement becomes slightly faster. --- 23–27 SEC — FINALLY The woman stops beneath the warm lights of an old cinema entrance. She realizes her scarf is missing and turns around. The man appears several meters behind her. He slowly raises the scarf. Their eyes meet for the first time. Everything around them continues moving normally, but the camera movement becomes calm and intimate. Neither speaks. She gives a surprised little smile. He smiles back. --- 27–30 SEC — FINAL FILMY SHOT He walks toward her and gently offers the scarf back. She takes it. Their hands briefly meet naturally while exchanging it. Both smile with slightly awkward, genuine expressions. Camera slowly pulls backward into a wide shot. Rain begins lightly again. People and cars move through the glowing city behind them. The two remain standing beneath the cinema lights talking quietly as the camera continues pulling away. CUT TO BLACK. --- CAMERA STYLE Premium romantic movie cinematography. ARRI Alexa 35 / Sony Venice cinematic realism.

Cinematic scene

Video Prompt: "Just Found My New Favorite Korean Lunch Spot" Style: Ultra-photorealistic 4K smartphone vlog, 7-shot continuous montage. Authentic handheld phone footage with subtle natural shake, autofocus breathing, realistic exposure shifts, HDR lighting, shallow depth of field, soft cinematic film grain. Feels like an organic travel vlog filmed by a friend—no staged or commercial look. A woman (reference image provides only her face and hairstyle) sits alone at a cozy Korean neighborhood restaurant beside a large open window overlooking a quiet alley. She wears a relaxed cream linen short-sleeve blouse with shoulders fully covered and light blue jeans. She remains seated on the same side facing the camera throughout the entire video. No other faces appear; only a server's hands briefly enter the frame. The wooden table holds a bubbling stone bowl of kimchi jjigae, a bowl of steamed rice, grilled seaweed, assorted banchan (kimchi, spinach, bean sprouts), stainless spoon and chopsticks, a glass of iced barley tea, and a folded napkin. Warm steam rises naturally while sunlight filters through the window, creating a cozy late-summer atmosphere. Shots 1. Arrival The steaming kimchi jjigae is placed on the table. Steam rises toward the camera, briefly fogging the lens. She smiles with genuine excitement and quietly says: "와... 냄새 미쳤다." 2. Mixing Overhead handheld shot. She scoops fluffy rice into the bubbling stew, gently mixing tofu, kimchi, and broth. Close-up of the rich red soup simmering. 3. First Bite She carefully blows on a spoonful, tastes it, pauses for a second, smiles naturally, and says: "진짜 맛있다." 4. Perfect Bite She wraps rice and kimchi in roasted seaweed, takes one satisfying bite, then laughs softly: "이 조합 최고다." 5. Refreshing Break She pours iced barley tea into a stainless cup. Condensation glistens on the glass. She drinks, exhales comfortably, and smiles. 6. Favorite Moment She lifts a large piece of silky tofu dripping with spicy broth, tastes it slowly, nods with approval, and says: "국물이 끝내준다." 7. Ending The meal is finished. Empty bowls remain on the table. She leans back, closes her eyes for a moment as a gentle breeze moves the curtains, then looks into the camera with a satisfied smile: "또 올게." Audio Only natural Korean dialogue where specified. Authentic ambient sounds only: Bubbling stew Spoon and chopsticks on ceramic Steam sizzling Rice mixing Tea pouring Light restaurant chatter (indistinct) Birds outside Distant city ambience Gentle breeze through the open window Negative Prompt No music, subtitles, captions, logos, watermarks, duplicated reference image, face morphing, identity changes, unrealistic food physics, exaggerated expressions, over-smooth skin, AI artifacts, floating objects, extra people, camera glitches, jump cuts, fast zooms, or cinematic transitions. Maintain consistent facial identity from the reference image, accurate lip-sync, natural blinking, realistic eating motions, and lifelike food textures throughout.

Fashion portrait

一句话了解

这个模型怎么样?

本页介绍 Kyeo 的 Runway 生成工作流。基础生成从文字或单图开始,Extend 则对符合资格的历史结果另开续写任务;它们有各自的输入条件和预扣费用。

模型定位
适合从文字或单图开始创作。若要改写已有视频,可了解 Runway Aleph 或 Luma Modify Video,并分别核对其输入条件和当前可用状态。
适合谁用
适合先生成一段短片,再判断是否需要续写。需要多图输入、现有视频改写或 1080p 10 秒生成时,应先核对其他工作流的实际可用能力。
本站使用说明
生成参数、组合预扣与符合条件的结果续写集中在同一工作流程中。提交前核对预计积分,完成后再按实际结算和资源状态安排保存。

关键参数一览

根据这些信息,判断模型是否适合你的任务。

生成方式
可从文字或单张图片开始生成视频;图生也需填写提示词。分辨率与时长按有效组合选择。
费用与结算
生成先按所选组合预扣积分;成功且实际消耗已确认、符合结算条件后,才结算并按规则退还差额。异常消耗继续待核。
提示词
必填,最多 1,800 字符;按工作台计数,部分表情会占用多个字符。
图片与画幅
最多 1 张 JPG/JPEG/PNG,不超过 10 MiB。文字模式可选 5 种比例;图生不另提交比例,不能据此保证成片沿用原图画幅。
Extend 续写
符合条件的成功 720p 结果可另开续写任务:输出 720p 预扣 12 积分,输出 1080p 预扣 30 积分。还需满足来源归属、任务标识与资源条件,不承诺新增时长。

它还叫什么

同一个模型在公开资料中可能有不同名称,这里整理相关称呼,方便查找和比较。

Runway
Runway 视频生成
Runway Extend

用户常见问题

从任务、输入条件和结果要求出发,整理选择模型时需要确认的问题。

Runway 怎么用
Runway 文生视频
Runway 图生视频
Runway 价格
Runway 720p 1080p 区别
Runway 5秒 10秒 区别
Runway 支持 1080p 10秒吗
Runway Extend 是什么
Runway Extend 怎么用
Kyeo 的 Runway 对应哪个官方版本

选型指南

根据输入素材、任务要求和积分成本,逐项确认模型是否适合。

1
从文字或单图创作

先确定要生成的画面和运动,再选有效组合。图片模式只使用单张起始图,仍需填写提示词。

2
改写已有视频

本页不把视频作为基础输入。改写已有视频时,可分别查看 Runway Aleph、Luma Modify Video 的输入要求及当前状态,再选择可用流程。

3
先核对组合与费用

默认 720p、5 秒;如需更长片段或更高分辨率,先核对有效组合和预扣额。Extend 另行收费,且来源资格不由一次成功生成自动保证。

模型对比

把常见候选放在同一任务下比较,帮助你确认输入方式、控制项和成本差异。

Runway 和 Runway Aleph 有什么区别
Runway 与 Runway Aleph 如何选择
Runway 和 Luma Modify Video 有什么区别
Runway 与 Luma Modify Video 如何选择

模型横向对比表

按同一组维度比较当前模型和候选模型的输入、设置与使用边界。

起始素材
Runway
从文字或单张图片开始;图生也需提示词,不接收现有视频作为本页生成输入。
Runway Aleph
以 1 个现有视频为必需输入,可另加 1 张参考图;具体条件按 Aleph 独立页面核对。
结果续写
Runway
Extend 是另一次任务,只对符合完整资格的成功 720p 来源提供;输出分辨率决定预扣额。
Runway Aleph
用于改写现有视频,不提供本页的 Runway Extend 动作。
控制方式
Runway
文字模式可选择比例,图生不另提交比例;生成和 Extend 分别提交、分别预扣。
Runway Aleph
围绕现有视频设置比例、水印文字和种子值;遵循独立的输入与计费条件。

适用情况

依据公开资料与当前站内设置,说明模型适用的任务及需要注意的限制。

本站工作流与官方型号分开核对

Runway 官方目录包含不同型号,不能据品牌名推定本站接入的具体版本。本页介绍 Kyeo 接入流程中的输入、组合和续写条件,版本对应关系仍待核对。

生成与续写分别预扣

基础生成按时长和分辨率组合估算;Extend 按输出分辨率估算。它们是两次独立任务,成功后的实际消耗确认与差额退还仍需分别满足结算条件。

能力详解

文字或单图起步

文字模式可选择画幅;单图模式不另提交比例。省略比例不保证保留原图方向、尺寸或构图。

明确的生成组合

720p 可选 5 秒或 10 秒,1080p 仅支持 5 秒;分辨率与时长不能任意交叉组合。

有条件的结果续写

成功 720p 只是来源资格的一部分。还需核对任务归属、标识、来源时长与可用资源,且不承诺新增片段长度。

适合做什么

从文字生成短片

用提示词说明主体、运动和画面重点,再按有效组合提交。无需为了文字模式额外上传图片。

从单图安排运动

先准备符合格式和大小要求的起始图,再描述希望出现的运动。输出比例与构图仍需在结果中核对。

为合格结果继续创作

在可用且符合来源条件的 Runway 结果上使用 Extend,单独写续写提示词并选择输出分辨率;不要据来源时长推算新增秒数。

提示词技巧

先定组合,再写画面

选择时长和分辨率后,集中描述这段片子的主体、运动和场景变化。1080p 不支持 10 秒,避免按不存在的组合安排内容。

单图也要描述运动

图片模式仍需非空提示词。上传前确认 JPG/JPEG/PNG 和大小限制,描述希望发生的动作,不把省略比例当作构图保留保证。

续写提示词另写

Extend 是独立请求,应描述希望后续出现的内容,并重新核对输出分辨率和预扣额。资格成立也不代表新增长度已有保证。

为什么选它

文字与单图的起始方式清楚,便于按已有素材选择。
生成组合逐项列出,便于提交前比较预扣额。
单图使用统一的格式、大小和上传校验要求。
合格结果可单独续写,续写输入与预扣另行核对。

使用前要知道

本页不接收多图或现有视频改写输入;其他工作流需另核其条件与可用状态。
不提供 1080p 10 秒组合;预扣额也不等于所有任务的固定最终消耗。
图片模式只接收符合 JPG/JPEG/PNG、大小与元数据校验要求的单图,不保证输出沿用原图画幅。
只有符合完整条件的成功 720p 来源才能续写;不承诺新增时长,1080p 来源不支持 Extend。
文件是否可播放、下载,以任务媒体的到期信息和当前状态为准。标记为 14 天保留的任务也应及时保存文件,不将保留策略视为媒体链接的可用保证。

可调参数

根据这些信息,判断模型是否适合你的任务。

视频时长
可选
参数类型: 选择
默认值:5 秒
5 秒
10 秒
视频分辨率
可选
参数类型: 选择
默认值:720p
720p
1080p
文字模式宽高比
可选
参数类型: 选择
默认值:16:9
16:9
4:3
1:1
3:4
9:16
水印文字
可选
参数类型: 文本

积分消耗

生成预扣:12 或 30 积分,依所选组合确定

基础生成有 3 种有效组合:720p 5 秒预扣 12 积分,720p 10 秒预扣 30 积分,1080p 5 秒预扣 30 积分。时长可选 5 或 10 秒,其中 1080p 仅支持 5 秒,不提供 1080p 10 秒。成功后的结算和差额退还需满足实际消耗确认条件,详见费用问答。

使用建议

准备多次生成时,建议先用少量素材测试,核对输入设置和结果是否适合任务,再决定是否继续。每次生成均按工作台显示的计费规则扣点。

常见问题

相关模型

按你的任务比较这些模型的输入要求、功能和当前状态,再查看对应详情。

Seedance 2.5

用 Seedance 2.5 规划较长镜头,先选择文字、首帧与首尾帧或多模态参考场景。请求时长可设为最长 30 秒;默认 5 秒,选择自动时长则按 30 秒预留预计积分。准备素材时,数量、单段时长与累计时长需分别核对。

AI 视频模型
预计预扣 102–5,700 积分

Seedance 2.0

从文字描述、首帧或参考素材出发制作视频。标准版提供文字、首尾帧和多模态参考场景;先选与素材相符的路径,再核对分辨率、预计积分和结果中需要检查的内容。

AI 视频模型
预计预扣 164–3,840 积分/次

Seedance 2.0 Fast

用文字规划镜头,也可从首帧或多模态素材出发生成视频。可选 480p、720p 和 4–15 秒整数时长;参考视频会影响预扣积分。

AI 视频模型
预计预扣 41–450 积分/次(含参考视频组合)

信息来源

页面内容依据公开模型资料与 Kyeo AI 的模型信息整理。能否在线使用,请以当前模型状态为准。

来源说明

本页按 Kyeo 接入的 Runway 流程说明输入、预扣与续写条件。官方模型目录仅供了解厂商型号;本站接入流程对应的原厂版本仍待核对。

内容复核日期: 2026-08-28
Runway:当前 API 模型目录
官方
查看原始来源