在 DomoAI 平台选用 Seedance 2.5 作为视频生成工具与模型。 原文：「Seedance 2.5 on @DomoAI_」
在提示词开头写入标题与总体设定，并指定视频时长为 10 秒。 原文：「TITLE: POV — DRIVING IN 2050 Create a 10-second ultra-photorealistic futuristic video showing a beautiful young woman casually driving along a highway in the year 2050, while hundreds of autonomous flying vehicles travel overhead in perfectly organized aerial traffic lanes.」
写入主角身份一致性描述，固定年龄、发型、妆容、服饰与配饰。 原文：「MAIN CHARACTER — STRICT IDENTITY CONSISTENCY A beautiful 23-year-old woman with long, straight pastel-pink hair, delicate facial features, realistic fair skin with visible natural texture, soft pink makeup, and subtle glossy lips. She wears: A fitted blue denim sleeveless top. A futuristic metallic silver choker necklace. Bright blue statement earrings. A black seatbelt across her chest. Maintain her exact face, hairstyle, outfit, proportions, and accessories throughout.」
写入场景 1 与场景 2 的分镜描述，先以车内 dashboard 机位拍摄驾驶者，再硬切到未来高速公路的空中交通。 原文：「SCENE 1 — 0:00–0:02 | INSIDE THE CAR Start with a fixed dashboard-mounted smartphone camera facing the woman from the passenger side. She sits behind the steering wheel of a modern luxury vehicle, driving casually. Warm golden-hour sunlight streams through the side window, illuminating her pink hair and face. She looks ahead at the road, then briefly glances toward the camera with a subtle, amused smile. The vehicle interior is realistic, with black leather seats, a dark roof, and natural windshield reflections. SCENE 2 — 0:02–0:04 | THE FUTURISTIC HIGHWAY REVEAL Hard cut to a spectacular elevated wide-angle view of a massive multilane highway at sunset. A dark luxury SUV drives along the nearly empty highway. Above the highway, hundreds of futuristic autonomous flying vehicles hover and travel in tightly organized aerial traffic lanes. The flying vehicles have compact aerodynamic bodies, metallic silver and dark-gray surfaces, and advanced concealed propulsion systems. They move smoothly in synchronized formations, maintaining precise distances from one another. Some fly directly above the highway while others stretch toward the horizon, creating a dense futuristic transportation network. Golden sunlight reflects off their metallic surfaces.」
写入场景 3 至场景 5 的分镜描述，回到驾驶者反应、展示大规模空中交通，并以最终微笑收尾。 原文：「SCENE 3 — 0:04–0:06 | BACK TO THE DRIVER Hard cut back to the same dashboard camera angle. The pink-haired woman continues driving completely unfazed by the extraordinary traffic system outside. She rests one hand naturally on the steering wheel and looks toward the windshield with a playful, slightly impressed expression. Sunlight shifts across her face as the vehicle moves. Her facial movements are subtle, natural, and believable. SCENE 4 — 0:06–0:08 | MASSIVE AERIAL TRAFFIC Cut to a dramatic elevated tracking shot following the dark SUV along the highway. The camera reveals an enormous formation of flying vehicles stretching across the sky. Several vehicles glide directly above the camera while hundreds more move across multiple aerial levels. The futuristic traffic looks technologically advanced but physically believable. Below, the dark SUV continues along the sunlit road, appearing tiny beneath the immense aerial transportation network. Warm orange sunlight, long road shadows, atmospheric haze, and realistic metallic reflections create a spectacular cinematic atmosphere. SCENE 5 — 0:08–0:10 | FINAL REACTION Cut back inside the car. The same pink-haired woman looks toward the camera and gives a small, knowing smile before returning her attention to the road. Golden sunlight illuminates her face. End on her calm, confident expression.」
写入镜头与视觉风格说明，指定车内手机 dashboard 质感、航拍镜头、手持微颤、自然光影与暖调日落调色。 原文：「CAMERA AND VISUAL STYLE Authentic smartphone dashboard-camera footage for interior shots. Spectacular aerial cinematography for exterior shots. Realistic handheld micro-vibrations inside the moving car. Natural sunlight, believable shadows, and atmospheric depth. Photorealistic flying vehicles with convincing scale, movement, and reflections. Smooth aerial tracking with realistic motion blur. Warm sunset color palette. Natural skin texture, visible pores, realistic hair movement. Strong contrast between ordinary everyday driving and extraordinary futuristic transportation.」
写入画面文字与音频说明，要求全程显示 “2050” 字样，并加入道路声、推进器嗡鸣与电子配乐。 原文：「ON-SCREEN TEXT Small, simple white text reading "2050" positioned near the top center throughout the video. AUDIO Natural road noise, soft futuristic electric propulsion hums, subtle wind sounds, and a dreamy futuristic electronic soundtrack.」
写入反向提示词与最终目标，排除卡通、动漫、游戏感、变脸、穿帮等问题，并强调视频要像 2050 年的真实病毒式社交片段。 原文：「NEGATIVE PROMPT No cartoon rendering, no anime, no video-game graphics, no artificial CGI-looking humans, no face morphing, no changing hair color, no inconsistent outfits, no distorted steering wheel, no extra fingers, no unrealistic flying vehicle collisions, no chaotic flight patterns, no flickering objects, no unnatural facial expressions, no excessive neon lighting, no dark nighttime setting. FINAL GOAL Make the video feel like an authentic viral social media clip recorded in the year 2050, where flying cars have become so ordinary that the woman barely reacts to hundreds of them passing overhead.」 未给项：工具平台=Seedance 2.5 on @DomoAI_；模型版本=2.5；时长=10-second；画幅=原帖未给；分辨率=原帖未给；参考素材=原帖未给
未给项：工具平台=Seedance 2.5 on @DomoAI_；模型版本=2.5；时长=10-second；画幅=原帖未给；分辨率=原帖未给；参考素材=原帖未给
在 Wizstar 平台上使用 Seedance 2.0 创建 AI 视频 原文：「Create it on Wizstar with Seedance 2.0」
上传图片 原文：「Upload your images.」
编写英文提示词，描述 Image 1 中的人物手持 Image 2 的产品进行展示，并要求保留产品原貌、生成自然英语对话 原文：「“ A confident, attractive woman from Image 1 holds the product from Image 2 for a presentation. Preserve the product’s exact appearance, design, colors, and details without alterations. Generate natural English dialogue suited to the scenario, with realistic expressions and smooth movements.”」
粘贴提示词并让 AI 生成视频 原文：「Paste the prompt. Let AI bring your product to life.」 未给项：工具平台=Wizstar；模型版本=Seedance 2.0；时长=原帖未给；画幅=原帖未给；分辨率=原帖未给；参考素材=Image 1、Image 2
在 Topview AI 平台输入提示词（该平台支持最多 1 万字符）。 原文：「プラットフォームは１万文字までプロンプト打てるTopview AI @TopviewAIJP」
将视频时长与拍摄形式设定为 15 秒、iPhone 手持连续拍摄的街头舞蹈视频。 原文：「A high-energy 15-second continuous iPhone-shot street dance music video on a busy modern Japanese city street.」
写入主体提示词，描述主舞者身份、舞蹈风格及“边跳边走”的核心要求。 原文：「A stylish female singer-dancer walks continuously toward the camera while performing energetic contemporary choreography combining modern hip-hop, waacking, K-pop dance, commercial street dance, body isolations, shoulder grooves, chest hits, hip accents, arm waves, quick footwork, and expressive performance. She is both a singer and dancer. She genuinely enjoys the music, occasionally mouths the lyrics, smiles naturally, makes confident eye contact with the camera, and performs with strong musicality and charisma. IMPORTANT: The dancer must continuously DANCE while WALKING. Do not reduce the choreography to ordinary walking with occasional arm movements. Her feet continuously travel forward while her entire body actively responds to the music. The choreography should have HIGH MOVEMENT DENSITY throughout the entire 15 seconds. The flash mob must begin VERY EARLY and gradually grow throughout the shot.」
加入摄影机设定提示词，限定 iPhone 手持自然后退跟拍，并禁止剪辑、变焦、无人机、慢动作等。 原文：「CAMERA: Shot entirely on a modern iPhone. Natural handheld smartphone cinematography. Realistic iPhone image quality. Sharp but natural smartphone detail. Slight natural handheld movement. Subtle smartphone stabilization. Natural exposure changes. Realistic autofocus behavior. Natural wide-angle perspective. No cinematic crane. No professional cinema camera. No gimbal-perfect movement. No dramatic lens effects. The camera operator continuously BACKTRACKS in front of the dancer while filming her. The camera moves backward along the city street while keeping the main dancer centered. The camera operator walks backward naturally but carefully. The main dancer remains full-body or near-full-body in frame. No cuts. No transitions. No orbiting. No camera rotation around the dancer. No zoom. No drone. No slow motion.」
按 0.0–15.0 秒的时间轴分段写入主舞者与群舞的详细动作编排。 原文：「0.0-1.5s | SOLO The scene begins immediately on a busy modern Japanese city street. The main dancer is already walking toward the camera and dancing. 0.0-0.5s: Right foot forward. Strong hip-hop bounce through the knees. Sharp shoulder hit. 0.5-1.0s: Left foot forward. Chest isolation followed by a shoulder roll. Her arms naturally swing into a short waacking movement. 1.0-1.5s: Right foot forward. Fast arm circle around the head. Hip accent. Small head nod toward the beat. She is clearly dancing, not simply walking. Several ordinary pedestrians are visible in the background.」
写入快闪人群递增规则与舞蹈密度要求，要求舞者数量随时间自然增加，并保持全身持续活跃。 原文：「FLASH MOB PROGRESSION This progression is extremely important: 0-1.5s: 1 dancer. 1.5-3s: 1 additional dancer begins joining. 3-5s: 2-3 additional dancers join. 5-7s: Several more dancers join. 7-9s: Approximately 8-10 dancers. 9-12s: More dancers appear naturally from the surrounding pedestrian flow. 12-15s: A lively moving street dance group surrounds the main dancer. The number of dancers should visibly increase over time. Do NOT suddenly spawn a large crowd. Do NOT remove dancers once they have joined. Do NOT teleport dancers. Each dancer must enter naturally from an existing position in the street environment. Some pedestrians remain ordinary pedestrians and never dance. This contrast makes the flash mob feel spontaneous and believable. ================================================== DANCE PRIORITY HIGH DANCE DENSITY. The main dancer should perform multiple coordinated movements every second. Walking is integrated into the choreography. Every forward step can contain: hip movement, shoulder groove, chest isolation, arm movement, head movement, or rhythmic footwork. The performer should look like a professional K-pop / hip-hop music-video dancer performing live on the street. Avoid simple walking. Avoid standing still. Avoid long poses. Avoid slow choreography. Keep the body active continuously.」
写入物理真实感、地点与最终风格提示词，指定现代日本城市街道与 iPhone 街拍氛围。 原文：「PHYSICAL REALISM Realistic human biomechanics. Clear weight transfer between legs. Natural knee flexion. Realistic hip movement. Feet remain grounded. Natural arm and shoulder mechanics. Natural torso rotation. No impossible joint angles. No foot sliding. No floating. No extra limbs. No duplicated body parts. No body deformation. ================================================== LOCATION Modern Japanese urban street. Wide sidewalk and road. Modern buildings. Storefronts. Traffic. Pedestrians. Bicycles and everyday city details in the background. The environment remains realistic and populated but never overwhelms the dancers. ================================================== FINAL STYLE Authentic iPhone-shot street dance video. Energetic. Spontaneous. Modern. Youthful. Musical. Urban. K-pop music-video energy mixed with real street-performance atmosphere. It should feel like someone happened to capture an incredible spontaneous dance sequence on an iPhone while walking backward through the city. 15 seconds total. ONE CONTINUOUS SHOT. CONTINUOUS BACKTRACKING CAMERA. HIGH DANCE DENSITY. EARLY FLASH MOB. GRADUAL INCREASE IN DANCERS.」 未给项：工具平台=Topview AI @TopviewAIJP；模型版本=Seedance2.5；时长=15 seconds total；画幅=原帖未给；分辨率=原帖未给；参考素材=原帖未给
未给项：工具平台=Topview AI @TopviewAIJP；模型版本=Seedance2.5；时长=15 seconds total；画幅=原帖未给；分辨率=原帖未给；参考素材=原帖未给
进入 Video 标签页，选择 Seedance 2 模型。 原文：「Here, we clicked into the video tab. We've chosen the seedance 2 model from the list, and we've chosen three reference images here.」
悬停上传区，添加图片、视频或音频作为参考素材。 原文：「We do that by hovering over here, and we can choose images or videos or audio files to add as references.」
在提示词中点击 @ 符号引用已上传的参考文件。 原文：「Once we've added them, we can refer to them in our prompt by clicking the at symbol.」
编写提示词，用 image / video / audio 编号指代对应参考，并描述需要使用的元素、场景或动作。 原文：「So, the prompt is that the man in image one is walking down the street from the video wearing the overalls from image three, and he's carrying the shirt from image four, and he's talking to the camera.」
若用音频做口型同步，把实际台词文稿也写进提示词以提升准确度。 原文：「anytime you're doing lip sync, if you can provide the actual transcript along with the audio file, you're bound to get much better results.」
若用视频参考，可在提示词中指定它作为角色动作或摄像机运动的参考。 原文：「use this video as a character motion reference for image one, which is that dog, riding a hoverboard in a futuristic city carrying an armful of tennis balls.」 未给项：工具平台=OpenArt；模型版本=Seedance 2.0（seedance 2）；时长=原帖未给；画幅=原帖未给；分辨率=原帖未给；参考素材=图片、视频、音频文件
在 YouArt 的 SeaDance 2 入口中，根据队列与写实角色表支持情况选择合适端点；处理写实角色表时可优先使用直连 Dreamina 的新端点。 原文：「So, I'm using SeaDance 2 on YouArt and now they have two different SeaDance 2 models. The first one is the legacy SeaDance 2 and Omni models. The other one is an endpoint that connects directly to the Dreamina platform. This new endpoint has a shorter queue and it works with realistic looking character sheets.」
进入 Nano Banana 2，上传一张带写实面部的原始角色表作为参考素材。 原文：「go to Nano Banana 2 and upload your original character sheet with a realistic face」
在 Nano Banana 2 中使用指定提示词，把角色表转换为轻微风格化的游戏角色渲染，同时保留面部识别度、面部特征、比例与服装设计等关键信息。 原文：「Convert this character sheet into a slightly stylized game character render. Keep the face highly recognizable, preserve facial features, bone structure, realistic proportions, and subtle skin detail. Reduce photorealism slightly. Use cleaner shapes, polished game-art rendering, and slightly simplified textures. Keep the original composition, pose, outfit, and character design unchanged. High-quality modern AAA video game character art.」
将 Nano Banana 2 生成的轻微风格化角色表作为参考图用于 Seedance 视频生成，并在视频提示词中加入真实感关键词，使最终输出仍保持写实。 原文：「While generating your videos, you also need to add these keywords to your prompts. Photorealistic live action, lifelike, and realistic skin texture.」
若仍触发平台过滤，可进一步让角色表更像游戏角色渲染，随后把它导入 legacy SeaDance 端点作为参考图，并加入 photorealistic live action 与 lifelike 关键词再生成。 原文：「Okay, let's say you did this method, but you are still getting errors. You can go one step further and make your character sheet a little bit more like a game character render. While generating this new character sheet, I used this prompt you see on the screen and then I took this character sheet and went to the legacy SeaDance endpoint. I attached it as my reference and added the keywords photorealistic live action and lifelike.」 未给项：工具平台=YouArt、Nano Banana 2；模型版本=Seedance 2.0、Seedance 2.0 Stable、legacy SeaDance 2、Omni models、Dreamina endpoint；时长=原帖未给；画幅=原帖未给；分辨率=原帖未给；参考素材=original character sheet with a realistic face
未给项：工具平台=YouArt、Nano Banana 2；模型版本=Seedance 2.0、Seedance 2.0 Stable、legacy SeaDance 2、Omni models、Dreamina endpoint；时长=原帖未给；画幅=原帖未给；分辨率=原帖未给；参考素材=original character sheet with a realistic face
实例
看原帖 / 原文 · 经验来源：YouTube DqWmHvIIcww · Bypass Face Filter