基于以下元提示词格式生成：{ "meta_prompt": { "description": "This meta prompt serves as a universal template for generating detailed Veo 3 prompts based on a user-provided concise scene description. The generated prompt must be comprehensive, covering single-character, dual-character, or multi-character scenes, and include details for lighting, object motion, camera techniques, camera movement, color tones, character voices, and environment. The output must be in JSON format, with both English and Chinese descriptions, and adhere to Veo 3 constraints (e.g., no subtitles, avoid backlighting unless specified, max 4 lines of dialogue per scene). The prompt must incorporate user-specified tone, style, and elements, and address known issues like audio desync or lighting hallucinations by including explicit instructions. If the user specifies a selfie video, include 'A selfie video of...', 'holds the camera at arm’s length', and 'occasionally looking into the camera'. For multi-shot scenes, split into sequences with clear timing. Ensure two versions per scene for iteration.", "structure": { "sequences": [ { "start_sec": "0", "end_sec": "8", "narrative": { "en": "{Detailed narrative description of the scene based on user input, capturing the main action or transformation, e.g., a character speaking, an object transforming, or environment shifting.}", "zh": "{基于用户输入的详细场景叙述，捕捉主要动作或转变，例如角色对话、物体变化或环境转换。}" }, "visuals": { "camera_setup": "{Cinematic or selfie setup, e.g., 'Cinematic, 50mm lens, f/1.8' or 'Selfie video, phone camera, arm’s length, arm visible'.}", "motion": "{Camera movement, e.g., 'eye level', 'high angle', 'worm’s eye', 'dolly shot', 'zoom shot', 'pan shot', 'tracking shot', or 'static'. Specify if selfie video includes 'occasionally looking into the camera'.}", "lighting": "{Lighting style, e.g., 'golden hour', 'neon glow', 'soft ethereal', 'sunset to night'. Avoid backlighting unless specified, prevent lighting hallucinations by defining light sources.}", "effects": ["{Visual effects, e.g., 'particle simulations', 'lens flares', 'mist particles', 'sand cascades', 'explosion of elements'.}"] }, "environment": { "setting": { "en": "{Location, e.g., 'Santa Monica Pier', 'retro-futuristic starport', 'enchanted forest', 'urban highway'.}", "zh": "{地点，例如‘圣莫尼卡码头’、‘复古未来星港’、‘魔法森林’、‘城市高速公路’。}" }, "atmosphere": { "en": "{Mood and environmental details, e.g., 'bustling with neon lights', 'misty and serene', 'stormy with crashing waves'.}", "zh": "{氛围和环境细节，例如‘霓虹灯闪烁的喧嚣’、‘薄雾弥漫的宁静’、‘波涛汹涌的风暴’。}" }, "props": ["{Objects in the scene, e.g., 'ferris wheel', 'hyperdrive lever', 'DJ booth', 'oasis palms'.}"] }, "entities": [ { "type": "{Character, object, or text, e.g., 'character', 'text (VEO3 title)', 'Corona bottle'.}", "details": { "en": "{For characters: age, ethnicity, clothing, expression, e.g., '25yo Latino male, streetwear, confident'. For objects/text: appearance, e.g., 'silver VEO3 title, futuristic font'.}", "zh": "{角色：年龄、种族、服装、表情，例如‘25岁拉丁裔男性，街头服饰，自信’。物体/文本：外观，例如‘银色VEO3标题，未来字体’。}" }, "behaviors": ["{Actions, e.g., 'speaks with Aussie drawl', 'coalesces from mist', 'vibrates and pops'.}"], "interactions": "{Interactions with other elements, e.g., 'points to ship', 'mist clings to text', 'sand cascades around letters'.}" } ], "audio_layers": { "background": "{Background music or ambient sound, e.g., 'orchestral swell at -10 dB', 'ocean surge at -11 dB', 'faint spaceship hum'.}", "dialogue": [ { "character": "{Character name or role, e.g., 'Captain HaHaa', 'Barista', 'Girl'.}", "text": { "en": "{Dialogue text, max 4 lines per scene, e.g., 'Check this beauty out!'. Specify voice tone, e.g., 'Aussie drawl', 'Staten Island accent'.}", "zh": "{对话文本，每场景最多4行，例如‘看看这宝贝！’。指定声音语气，例如‘澳洲口音’、‘斯塔滕岛口音’。}" } } ], "effects": ["{Sound effects, e.g., 'whooshing particle builds', 'engine hum', 'sand shifts'.}"] } } ], "constraints": { "takes": "Generate 2 versions per scene to ensure quality iteration.", "dialogue_limit": "Max 4 lines of dialogue per scene to avoid audio desync. Split longer conversations into multiple sequences.", "no_subtitles": "No subtitles in the video.", "no_backlighting": "Avoid backlighting unless specified, use front/side lighting to prevent hallucinations.", "focus": "No blurred background, everything in focus unless specified.", "veo3_selection": "Explicitly select Veo 3 with experimental audio to avoid defaulting to Veo 2.", "dialogue_clarity": "Specify character roles clearly to prevent dialogue mixups (e.g., 'male asks, female answers').", "selfie_specifics": "If selfie video, include 'A selfie video of...', 'holds the camera at arm’s length, arm visible', and 'occasionally looking into the camera'." }, "output_format": { "version_1": { "en": "{Complete prompt in English, following the sequence structure above.}", "zh": "{完整的中文提示词，遵循上述序列结构。}" }, "version_2": { "en": "{Second variation of the prompt in English, with slight tweaks to camera, lighting, or dialogue for iteration.}", "zh": "{第二个中文提示词变体，略微调整摄像头、灯光或对话以进行迭代。}" } } }, "instructions": { "input_processing": "When the user provides a concise scene description, interpret the tone, setting, characters, and actions. Map these to the structure above, filling in details like camera movement (e.g., eye level, dolly shot), lighting, audio, and props based on the context. Use provided examples (e.g., Santa Monica Pier, Captain HaHaa, Corona bottle rave) to guide style and tone.", "error_handling": "Address Veo 3 bugs: prevent audio desync by limiting dialogue, avoid lighting hallucinations by specifying light sources, ensure Veo 3 selection, and clarify character dialogue roles.", "language": "Generate prompts in both English and Chinese, ensuring cultural and contextual accuracy in translations.", "examples": [ "Santa Monica Pier interview: Casual TikTok vibe, golden-hour lighting, static shot, male interviewer, female responder.", "Captain HaHaa starport: Retro-futuristic, cheeky Aussie tone, medium close-up to over-the-shoulder reveal.", "Corona bottle rave: Cinematic magical realism, close-up to crane shot, sunset to neon lighting.", "Selfie video: Arm’s length, occasional camera glances, dynamic urban or natural setting." ]