元峰AI笔记 · AI绘画与视频

实战案例:15组场景谁更胜一筹?

<title等了一年多,Midjourney的视频模型终于要来了! 刷到这个消息时,整个人都开始兴奋起来了</title

等了一年多,Midjourney的视频模型终于要来了! 刷到这个消息时,整个人都开始兴奋起来了

等了一年多,Midjourney的视频模型终于要来了!

刷到这个消息时,整个人都开始兴奋起来了。作为AI绘画界的扛把子,Midjourney终于要把视频这个大招憋出来了!

据可靠消息,Midjourney不仅要在本月推出视频生成功能,V7.1和V8模型也在马不停蹄地开发中。

这位"审美天花板"终于要进军视频领域了,凭借它在AI绘画上的绝对统治力,这波操作绝对会让整个创意圈为之震动!

图片展示的是Midjourney的“视频评分派对!第2轮”页面。页面左侧是Midjourney平台界面,右侧是活动说明。说明提到邀请用户对视频模型进行评估,在第2轮测试中发现模型在表现“去拿东西”等动作上有所进步,但仍有不足。还列出了参与评估的步骤,如按时观看视频、选择最喜欢的视频等。该图片与上文提到的Midjourney正在搞“视频评级派对”活动相呼应,展示了活动页面及相关说明。

他们正在搞一个"视频评级派对"活动,邀请用户参与评估优化最终模型。如果你是订阅用户,现在就可以访问体验: midjourney.com/rank-video

然后看到有人拿Midjourney跟当前市场上的两大视频生成巨头——谷歌的Veo3和快手的可灵2.1硬刚了一波。用完全相同的提示词,测试了12组不同场景。

先给大家扒一扒今天参与挑战这三位"选手"各自有啥看家本领:

Midjourney视频模型:Midjourney 作为知名的 AI 图像生成工具,其视频模型备受关注。根据大家的评价, **Midjourney 视频模型(目前主要指 V1 版本)的特点艺术感拉满,但在逼真度和物理效果方面还有点青涩。**它的视频功能将以图像到视频为核心,初期只对年度订阅用户开放,暂不支持文本到视频或外部图像输入。

Veo3:Google 推出的 Veo3 模型在 AI 视频生成领域被认为是里程碑式的突破,其在全网获得了极高 的评价。**音画同步生成,能让视频中的人物真正"开口说话",唇形都同步的那种。它不仅能生成高质量4K视频,还能自动生成与画面完美同步的对话、音效和背景音乐。**在篮球投篮等复杂场景中,连球落地、球鞋摩擦声都模拟得惟妙惟肖。

可灵2.1:快手的亲儿子,最近降价65%,最新的2.1模型性价比直接拉满,也是我每天都用的工具。2.1模型分三个版本:标准版(720P)、高品质版(1080P)和大师版(1080P)。动态表现和物理规律方面有明显提升,比如车辆落地时的物理反馈、人物动作的合理性都很到位,也可以生成音效。

实战案例:15组场景谁更胜一筹?

来看看12个场景中的几个典型案例:

俯视角度僵尸射击游戏

Veo3完胜!不仅画面流畅,连枪声、僵尸嘶吼和脚步声都精准同步,简直就是个小游戏demo。

Midjourney的画面风格很有特色,但动作有点卡顿,缺乏真实物理效果。

阳光下的公园足球场

A photorealistic scene in a sun-drenched park where a group of friends transitions from casual conversation to soccer enthusiasm. The camera gently pans across their expressive faces as attention shifts to the ball, which one friend nudges with his foot. Their body language becomes more animated as they discuss techniques, culminating in an impressive ball-spinning trick that draws admiring looks. Throughout, golden sunlight bathes the scene in warm tones, highlighting their camaraderie against the backdrop of vibrant green grass and leafy trees. 阳光明媚的公园里,一群朋友从轻松聊天转变为足球热情。镜头缓缓扫过他们充满表情的脸庞,注意力转向脚下的球,其中一个朋友用脚轻轻推了一下球。他们的肢体语言变得更加生动,讨论着技巧,最终展示了一个令人赞叹的球技,赢得了羡慕的目光。整个场景被金色的阳光温暖地照亮,金色的阳光与鲜艳的绿色草地和茂密的树木相互映衬,凸显出他们之间的亲密情谊。

战后雾气笼罩的村庄

A military jeep rolls into the foggy post-WWII village, parting the mist as it moves down the cobblestone street. A soldier jumps from the still-moving vehicle and rushes to knock urgently on a building door while windows open above as villagers peer out at the commotion. Autumn trees shake in a sudden gust, showering colorful leaves across the scene as the door opens, spilling golden light onto the wet stones. The jeep continues onward, its red taillights gradually disappearing into the thickening fog as the camera tilts upward to reveal the village rooftops. 一辆军用吉普车驶入战后的雾村,车轮卷起雾气,沿着鹅卵石街道前行。车上的士兵跳下车,急匆匆地敲响一扇房屋的门,与此同时,屋顶的窗户也打开了,村民们探头向外张望。一阵突然的秋风摇动着树木,五彩斑斓的落叶洒满了整个场景。门被打开,金色的光线洒在湿润的石板上。吉普车继续前行,红色的尾灯逐渐消失在越来越浓的雾中,镜头逐渐向上拉,露出雾村的屋顶。

神秘的兜帽人和发光的遗物

The video begins in a forgotten corner of an ancient structure where a solitary hooded figure sits on crumbling stone steps, with tendrils of vegetation creeping across weathered walls and a mysterious golden light glowing in the distance. The camera slowly pushes in as the figure stands and reaches into their cloak, withdrawing a small object that emits a soft blue glow between their fingers. Wisps of mist begin to curl around the steps, rising from below as if summoned. The distant golden light intensifies in response, causing the vines along the walls to cast moving shadows that seem to reach toward the figure. A raven swoops into frame, circling once before landing on a nearby stone railing, its attention fixed on the glowing object. The video ends with the figure holding the glowing object aloft, the mist now swirling more vigorously around their feet while the raven spreads its wings in apparent agitation, creating a tableau of mystical tension. 视频开始于一座古老建筑被遗忘的角落,一个戴着兜帽的身影坐在破败的石阶上,藤蔓沿着风化的墙壁蔓延,远处隐约散发着神秘的金色光芒。镜头缓缓推进,当身影站起并伸手入袍时,手指间发出柔和的蓝光。薄雾开始环绕石阶,从下方升起,仿佛被召唤而来。远处的金色光芒随之增强,使得墙壁上的藤蔓投下移动的阴影,似乎在向身影伸展。一只乌鸦飞入画面,绕圈一圈后落在附近的石栏杆上,目光紧盯着发光的物体。视频以身影高举发光物体结束,脚下雾气更加剧烈地翻腾,乌鸦展开翅膀显得焦躁不安,形成一种神秘的紧张氛围。

外星峡谷与尘埃

An alien desert canyon with warm ochre rock formations under a teal sky with multiple moons. The camera pans slowly across the landscape as a dust devil forms and spirals upward, catching sunlight on its swirling particles. Clouds reshape and cast moving shadows while a small lizard-like creature scurries across a rock. Stylized digital illustration with a painterly quality and sci-fi atmosphere. 一片温暖赭石色岩石形成的外星沙漠峡谷,天空湛蓝,有多个月亮。摄像机缓慢扫过这片景色,尘旋风形成并向上旋转,阳光在其旋涡状颗粒上闪烁。云朵不断变化,投下移动的阴影,一只小蜥蜴般的生物在岩石上爬行。风格化的数字插画,带有油画质感和科幻氛围。

花影下的战士

An anime warrior with fiery orange hair stands beneath cherry blossom trees, their hair billowing dramatically as they slowly look upward. A strengthening wind causes their dark uniform to flutter and the branches to sway, releasing a shower of pink petals that swirl around them as they raise one hand to feel the breeze. Rendered in vibrant Genshin Impact style with dramatic lighting that highlights the contrast between the warm orange hair and the delicate pink blossoms. 一位火红色长发的动漫战士站在樱花树下,她的长发随风飘扬,缓缓抬头望向天空。增强的风使她深色的制服飘动,树枝摇曳,粉色的花瓣如雨般飘落,围绕在她身边。她抬起一只手感受着微风。画面采用充满活力的原神风格,强烈的灯光突出了她温暖的火红色头发与精致的粉色樱花之间的对比。

幻梦中的能量轨迹跑者

The silhouetted figure runs steadily across the misty landscape, their form outlined by pulsing energy trails that brighten with each footfall. Their arms pump rhythmically as they move from left to right, creating subtle ripples of light on the dark surface beneath them. The teal-blue atmosphere and distant city lights remain constant as the runner continues their determined journey through the dreamlike, cinematic environment. 轮廓分明的身影在雾气缭绕的景色中稳步奔跑,每一步都伴随着闪烁的能量轨迹,使其身影在脚下逐渐明亮。他们的双臂有节奏地摆动,从左至右移动,在暗色的地面上创造出细微的光波。青蓝色的氛围和远处的城市灯光保持不变,跑步者在梦幻般的电影环境中继续坚定地前行。

穿越沙漠悬崖,尘土飞扬的吉普车

An epic vista of a rugged desert valley with towering cliffs and swirling dust. A rugged jeep speeds into the frame from the left, kicking up a large dust trail. The camera quickly tracks alongside the vehicle, panning rapidly to follow its movement towards the distant mountains. Cinematic, dramatic lighting. 一片崎岖的沙漠山谷,高耸的悬崖和旋涡状的尘土。一辆粗犷的吉普车从左方疾驰进入画面,扬起一大片尘土轨迹。摄像机迅速跟随着车辆,快速扫过以追踪其向远处山脉移动的轨迹。电影般的戏剧性光线。

坚定的表情

Hyper-realistic close-up of a contemplative face with natural freckles and texture, eyes slowly blinking then gazing upward, eyebrows subtly raising, lips parting slightly before pressing together, ending with a resolute expression as they stare into the distance against a neutral background. 超写实近景,一张充满深思的脸,自然分布着雀斑和纹理,眼睛缓缓眨动后凝视向上,眉毛微微扬起,嘴唇微微张开又合上,最终呈现出坚定的表情,背景中立。

穿白色网球装的女子

A white woman wearing a smooth white tennis dress on the street. she is on a tree-lined West Village sidewalk, looking at the camera. Her neat brown ponytail has a white scrunchie and subtle blonde highlights. This award-winning full body portrait demonstrates popularity beauty and confidence. She is pretty and athletic with an hourglass figure. 一位穿着光滑白色网球裙的白人女子站在街道上。她站在西村一条林荫大道的人行道上,正对着镜头微笑。她整齐的棕色马尾扎着一个白色发圈,还有一丝淡金色的发色。这幅获奖全身肖像展示了她的流行美感和自信。她身材匀称,既漂亮又健美。

两只猫在图案沙发上放松

In a photographic, realistic scene, an orange tabby and a tortoiseshell cat sit on a patterned couch. The tortoiseshell cat stretches out a paw and kneads the cushion gently. The orange cat shifts its weight, stands up slowly, and stretches its back before settling back down in a slightly different position on the couch. 在一张写实的照片中,一只橙色虎斑猫和一只虎斑猫坐在有图案的沙发上。虎斑猫伸出爪子轻轻揉着垫子。橙色猫移了移体重,慢慢站起来,伸了个懒腰,然后在沙发上换了个稍微不同的姿势躺下。

3D CGI 海滩场景置于玻璃球内

A short video clip in the style of 3D CGI animation, whimsical and cheerful, showing a miniature beach scene inside a glass sphere. A stylized woman holding an umbrella takes a few steps forward, a child turns their head, and the large man blinks. The camera subtly pans left to follow the woman. 一段风格化的 3D CGI 动画短视频,画面 whimsical 而愉快,展示了一个玻璃球内的迷你海滩场景。一个造型化的女士手持雨伞向前走了几步,一个孩子转头看了看,而那个大个子则眨了眨眼。镜头微微向左移动,跟随女士的视角。

超实用选择指南:哪款最适合你?

说了这么多,到底哪款最适合你?通过上面的案例分析,聊点我自己的的真实感受

预算党必选:可灵2.1

我懂,钱包永远是痛点。可灵2.1模型性价比真的无敌,1080P分辨率对大多数短视频平台完全够用,而且标准版价格便宜到让你怀疑人生。如果你是刚起步的创作者或者预算有限的小团队,这绝对是你的首选。

音频需求:只能是Veo3

这方面Veo3独孤求败,没有之一。**如果你需要人物说话、环境音效、背景音乐一条龙服务,别犹豫,Veo3是唯一选择。**尤其是做教育内容、产品演示或者需要角色对话的视频,Veo3简直是救星。缺点就是...贵,真的挺贵的。

图生视频:可灵2.1更胜一筹

手上有大量静态图片想转成视频?可灵2.1的图生视频能力目前是最强的。无论是产品展示、PPT动画化还是照片动起来,可灵2.1都能给你惊喜。而且它支持最长10秒视频,比其他两家都长。

艺术追求党:密切关注Midjourney

如果你是那种对美学有极致追求的创作者,尤其是已经深度使用Midjourney图像生成的用户,那就再等等。至于Midjourney,老实说,它的视频功能现在还不够成熟,但凭借其在图像领域的强大实力,未来肯定不容小觑。如果你不着急用,可以再等等看。V8模型一出,可能会带来惊喜。

未来,AI 视频生成技术将继续朝着更高质量、更长时长、更强交互性和更低成本的方向发 展。随着技术的不断成熟,AI 视频模型将深刻改变内容创作、影视制作、广告营销等多个行 业,为人类带来前所未有的视觉体验和创作自由。