Kling O3
当前状态 — 当前开放可完整验证的单镜头文字生成,以及单张首帧或首尾帧两图生成。 参考视频、视频转换、多镜头和元素工作流的合同分支仍有歧义,因此继续关闭。
快速了解
- 当前开放可完整验证的单镜头文字生成,以及单张首帧或首尾帧两图生成。 提示词必填且最多 3,072 字符;时长为 3–15 秒整数。 每次 42–1005 积分:720p 无/有音频为 14/18 积分/秒,1080p 为 18/23,4K 为 67。
- 每次 42–1005 积分:720p 无/有音频为 14/18 积分/秒,1080p 为 18/23,4K 为 67。 本地安全子集接受 1–2 张 JPEG/PNG 图片;单张不超过 10 MB,宽高均大于 340 px,比例为 0.4–2.5。
- 参考视频、视频转换、多镜头和元素工作流的合同分支仍有歧义,因此继续关闭。 当前开放可完整验证的单镜头文字生成,以及单张首帧或首尾帧两图生成。
相关查询: Kling O3 现在能用吗? · Kling O3 支持哪些视频模式? · Kling O3 有官方价格吗?
示例
Cinematic scene
A realistic Red Bull World Cup 2026 commercial scene with an elite freestyle footballer standing on the roof of a skyscraper overlooking a stadium city, holding a Red Bull can dripping with condensation - He opens the can, close-up on the crack and fizz; - Energy kicks in, he sprints across rooftops, juggling the ball, jumping gaps and performing impossible freestyle tricks as helicopters film him; - He volleys the ball from the rooftop toward a giant stadium screen, ending with the Red Bull can in foreground and the city celebrating below - Hyper-realistic sports commercial, golden daylight, drone tracking shots, dynamic slow motion, sweat, sparks, urban adrenaline, premium action advertising.
Cinematic scene
Ultra-realistic viral motocross freestyle spectacle with absurd premium sports-commercial energy. Shot on ARRI Alexa 35 with anamorphic lenses, combining low-angle hero shots, aggressive tracking, dramatic push-ins, wheel-level close-ups, speed ramps and suspended slow motion. Bright golden daylight, dust clouds, roaring crowd, fairground-sport atmosphere, premium live-action realism, no cartoon look. Main subject: clearly visible elderly grandmother motocross rider, deeply lined face, grey hair in a neat bun or tucked under a small floral headscarf, warm but fearless expression, no helmet, unmistakably old from the first frame. She wears a classic grandmother outfit: long floral dress, beige knitted cardigan, sensible stockings and old-fashioned shoes, creating a hilarious contrast with her unbelievable motocross skill. Background: excited spectators, one announcer, a few younger riders watching in disbelief. Outdoor motocross freestyle arena in daylight, dirt ramps, landing mounds, crowd barriers, open sky, dust and sunlight, banners without readable text. 0-5s: Immediate hero reveal Start with a clear medium close-up showing her wrinkled face, grey hair, floral dress and cardigan while she sits on a motocross bike at the start line. The audience is confused. Close-up on her calm grandmother face, hands on the handlebars, cardigan fluttering lightly. She revs the bike and launches forward. 5-10s: Dynamic tracking She accelerates hard toward the first dirt ramp. The camera tracks low beside her, emphasizing the absurd contrast between her grandmother outfit and the roaring bike. Dust explodes behind the wheels as she commits fully. 10-15s: Extreme slow-motion freestyle moment She hits the ramp and flies high into the air. In slow motion she performs a completely outrageous freestyle move while still clearly wearing her grandmother outfit: body stretched elegantly off the bike, one hand extended, dress and cardigan fluttering, then smoothly returning to control. Close-ups on the spinning wheels, her focused grandmother face, and the crowd in total disbelief. 15-20s: Epic viral finish She lands perfectly, charges straight into a second jump and performs an even crazier stylish aerial figure, then lands cleanly and rides past camera with total calm, still clearly a grandmother in full granny clothing. The crowd erupts. The announcer shouts in English: “That’s somebody’s grandma?!” Photorealistic motocross physics, grandmother clearly identifiable from the very first second, no helmet, full granny outfit visible throughout, realistic suspension compression, premium slow motion, stable rider continuity, strong viral contrast, clean landings, no crash, no injury, no readable text.
Cinematic scene
Create a 10-second cinematic travel vlog montage. Use the uploaded woman’s face consistently throughout every shot. She is a stunning young female travel creator with long dark wavy hair with curly bangs,wearing fashionable oversized Japanese streetwear (burgundy long leather jacket,shoulder bag), natural makeup, cheerful personality. The entire video should feel like real smartphone footage, with handheld camera movement, imperfect framing, accidental lens flare, slight motion blur, rolling shutter, high ISO noise at night, autofocus hunting, overexposed highlights, slight barrel distortion from a phone wide-angle lens, tiny lens smudges, and natural camera shake. No polished commercial look. Story Flow (10s) 0.0–0.8s Low-angle selfie in front of Tokyo Tower. She smiles naturally while adjusting the phone. Slightly tilted horizon. Bright overexposed cloudy sky. Tiny fingerprint on lens. 0.8–1.6s Walking through Shibuya Crossing at night. Looks back at camera while walking. Neon reflections on wet road. Strong motion blur. Quick whip-pan transition. 1.6–2.4s Inside a Japanese convenience store. Holding an onigiri close to camera. Playful eyebrow raise cutely . Harsh fluorescent lighting. Background shelves softly blurred. 2.4–3.2s Sitting alone in Yoyogi Park. Quiet candid moment. Looking away from camera. Soft sunlight filtering through trees. Slight focus miss. 3.2–4.0s Eating freshly made takoyaki. Instant reaction after first bite. Warm lantern lighting. Face slightly blurred from movement. Laughing naturally. 4.0–4.8s Standing in front of a glowing vending machine. Blue and pink lighting on face. Pressing drink button casually. Visible night grain. 4.8–5.6s Reflection in Tokyo tram window. Layered city lights reflected on glass. She looks outside thoughtfully. Train moving creates light streaks. 5.6–6.4s Walking beneath a Torii gate. Low-angle tracking shot from behind. Turns her head briefly. Motion blur. Bright halo from sky. 6.4–7.2s Immersive digital art exhibition. Blue, purple and gold lights moving across face. Looking upward in amazement. Natural low-light noise. 7.2–8.0s Wide-angle low-angle shot beside the giant silver bullet sculpture. She appears small beneath the sculpture. Phone lens distortion. Sky slightly blown out. 8.0–8.8s Walking through a lantern-lit alley. Camera follows behind. She turns around with a smile. Warm lantern glow. Natural night grain. 8.8–10.0s Rooftop selfie overlooking Tokyo skyline. Wind blows hair across camera lens. Slightly tilted horizon. Overexposed city lights. She laughs naturally while extending one arm toward the skyline. Editing Style * Fast whip-pan transitions * Speed ramps * Match cuts * Natural jump cuts * Handheld camera movement * Slight rolling shutter * Smartphone HDR exposure * Real travel vlog energy * No slow motion * No cinematic movie grading * Authentic TikTok creator aesthetic * Warm skin tones * High realism * Consistent face identity across every scene * 16:9 landscape
Editorial design
Visual transformation
一句话了解
这个模型怎么样?
每次 42–1005 积分:720p 无/有音频为 14/18 积分/秒,1080p 为 18/23,4K 为 67。 提示词必填且最多 3,072 字符;时长为 3–15 秒整数。
- 模型定位
- 当前开放可完整验证的单镜头文字生成,以及单张首帧或首尾帧两图生成。 提示词必填且最多 3,072 字符;时长为 3–15 秒整数。
- 适合谁用
- 提示词必填且最多 3,072 字符;时长为 3–15 秒整数。 每次 42–1005 积分:720p 无/有音频为 14/18 积分/秒,1080p 为 18/23,4K 为 67。
- 本站使用说明
- 本地安全子集接受 1–2 张 JPEG/PNG 图片;单张不超过 10 MB,宽高均大于 340 px,比例为 0.4–2.5。 参考视频、视频转换、多镜头和元素工作流的合同分支仍有歧义,因此继续关闭。
关键参数一览
根据这些信息,判断模型是否适合你的任务。
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。
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。
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。
现在需要 Kling 视频 — 当前开放可完整验证的单镜头文字生成,以及单张首帧或首尾帧两图生成。
准备全模态评测 — 参考视频、视频转换、多镜头和元素工作流的合同分支仍有歧义,因此继续关闭。
避免价格外推 — 每次 42–1005 积分:720p 无/有音频为 14/18 积分/秒,1080p 为 18/23,4K 为 67。
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。
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。
能力文档不等于可安全计费
全模态分支必须独立验收
能力详解
文字生成视频
图片与参考生成
视频转换
适合做什么
多镜头概念片
角色参考短片
视频风格转换
提示词技巧
按分支保存样本
记录素材约束
先设验收指标
为什么选它
使用前要知道
可调参数
根据这些信息,判断模型是否适合你的任务。
积分消耗
每次 42–1005 积分:720p 无/有音频为 14/18 积分/秒,1080p 为 18/23,4K 为 67。
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。
Kling 3.0
用文字或图片制作视频,可选择单镜头,也可逐镜头安排提示词与时长。声音和画质档位会参与预计预扣计算;人物、口型、文字、声音及镜头衔接都需检查实际结果。
Kling 3.0 Turbo
用文字或 1 张起始图制作短视频,选择分辨率与 3~15 秒整数时长。预计预扣由分辨率和请求时长计算;生成后仍需检查画面、素材使用权和实际结算状态。
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。
当前开放可完整验证的单镜头文字生成,以及单张首帧或首尾帧两图生成。 参考视频、视频转换、多镜头和元素工作流的合同分支仍有歧义,因此继续关闭。
每次 42–1005 积分:720p 无/有音频为 14/18 积分/秒,1080p 为 18/23,4K 为 67。 参考视频、视频转换、多镜头和元素工作流的合同分支仍有歧义,因此继续关闭。