AI агенти за генериране на видео

src:
  structure.yaml:
    content: |-
      Project: Creating Innovative Digital Art Fusing Surrealism and Futurism
      - Fantastical world of crystal islands floating on an ocean of liquid starlight
      - Surreal landscape with clockwork mechanical trees and luminous butterflies
      - Photorealistic yet dreamlike still image generation using t2i-imagen3
      - Magical motion and cosmic energy animation using Kling i2v v2.1
      - Unique visual expression where fantasy meets steampunk
      - Mystical space with gravity-defying upward waterfalls and multiple moons
      - Fantastical ecosystem of bioluminescent plants and cosmic energy flows
    agent_selection_reason: Selected claude-3-5-sonnet-20241022 for its excellence in managing project structure design and multiple media generation requirements. Particularly suited for planning surreal artworks and integrated multimedia production management.
    dependency: []
    agent: claude-3-5-sonnet-20241022
    api: []
  image-generation:
    beautifle_landscape.png:
      content: A surreal dreamscape where floating crystal islands hover above an ocean of liquid starlight. Ancient clockwork trees with gears for leaves stretch toward a sky filled with multiple moons in different phases. Ethereal butterflies made of pure light weave between cascading waterfalls that flow upward, defying gravity. The entire scene is painted in vibrant purples, electric blues, and golden amber hues, creating a mystical atmosphere that blends fantasy with steampunk aesthetics. Intricate details include floating geometric structures, bioluminescent flora, and streams of cosmic energy connecting the floating islands.
      agent_selection_reason: t2i-imagen3, Google's cutting-edge image generation model, excels at simultaneously expressing surreal elements and photorealistic textures. It particularly demonstrates advanced capabilities in convincingly depicting liquid starlight, luminous organisms, complex light interactions, and gravity-defying physical phenomena. It also shows superior ability in integrating the complex aesthetics of fantasy and steampunk fusion into a single work.
      dependency: []
      note: Prompts must be in English only.
      agent: t2i-imagen3
      api: []
  video-generation:
    beautifle_landscape.mp4:
      content: |-
        Transform this surreal dreamscape into a mesmerizing animated masterpiece.
        The crystal islands gently rotate and pulse with inner light while floating gracefully.
        Clockwork trees' gear-leaves turn slowly, creating hypnotic mechanical rhythms.
        Light-butterflies dance in spiraling patterns, leaving trails of stardust.
        The upward-flowing waterfalls shimmer and undulate with magical energy.
        Multiple moons slowly orbit across the cosmic sky, casting shifting shadows.
        Bioluminescent plants pulse in synchronized waves of color.
        Streams of cosmic energy flow between islands like living aurora.
        The liquid starlight ocean below ripples with celestial reflections.
        Camera slowly pans across this fantastical realm with cinematic grandeur.
      dependency:
        - src/image-generation/beautifle_landscape.png
      dependency_wait: true
      agent: i2v-kling-v2.1
      api: []
      parameters:
        duration: 5
        aspect_ratio: "1:1"
        negative_prompt: low quality, blurry, static camera
        cfg_scale: 0.6

src:
  structure.yaml:
    content: |-
      Project: Standard Quality Video Generation with Kling v2.1 Standard Edition
      - High-quality image-to-video conversion with standard settings
      - Multiple aspect ratio support (16:9, 9:16, 1:1)
      - Duration options (5 or 10 seconds)
      - Consistent and reliable video generation
    agent_selection_reason: Using i2v-kling-2.1-standard for stable and consistent video generation with standard settings
    agent: t2i-flux-schnell
    dependency: []
  image-generation:
    sample-image.png:
      content: |-
        A serene landscape with a calm lake reflecting mountains
      agent: t2i-flux-schnell
      dependency: []
      api: []
  video-generation:
    landscape-video.mp4:
      content: |-
        Gentle breeze creating ripples on the water surface
        Clouds slowly drifting across the sky
        Subtle light changes as time passes
        Natural and peaceful atmosphere
      dependency:
        - src/image-generation/sample-image.png
      dependency_wait: true
      agent: i2v-kling-2.1-standard
      api: []
      parameters:
        duration: 5
        aspect_ratio: "16:9"
        negative_prompt: low quality, unnatural movement
        cfg_scale: 0.5

src:
  structure.yaml:
    content: |-
      Project: Cinematic Orbit Shot of Futuristic Explorer
      - Single still image converted into a moving shot
      - Camera arcs around subject at sunset
      - Tail frame highlights silhouette against the sun
    agent_selection_reason: Selected i2v-kling-v2.5-turbo-pro for smooth cinematic motion.
    dependency:
      - src/local_image/explorer.png
    dependency_wait: true
    agent: i2v-kling-v2.5-turbo-pro
    api: []
    parameters:
      duration: "10"
      aspect_ratio: "16:9"
      negative_prompt: blur, distort, and low quality
      cfg_scale: 0.5

src:
  structure.yaml:
    content: |-
      Project: Cinematic Intro with Seedance Lite Reference-to-Video
      - Use multiple reference stills to control motion
      - Focus on neon-lit cyberpunk ambience and flowing garments
      - Track the subject with a gentle forward-moving camera
    agent_selection_reason: Using i2v-bytedance-seedance-v1-lite-reference-to-video for reference-driven motion control.
    agent: t2i-flux-schnell
    dependency: []
  image-generation:
    reference-main.png:
      content: |-
        A cyberpunk hero standing in a neon-lit alley, dramatic lighting on the coat.
      agent: t2i-flux-schnell
      dependency: []
      api: []
    reference-style.png:
      content: |-
        Stylized lighting accents and atmosphere for the same alley scene.
      agent: t2i-flux-schnell
      dependency: []
      api: []
  video-generation:
    hero-intro.mp4:
      content: |-
        Generate a cinematic opener where the protagonist walks through the neon-lit alley, coat flowing in the wind. Maintain a gentle tracking shot from a slightly elevated angle.
      dependency:
        - src/image-generation/reference-main.png
        - src/image-generation/reference-style.png
      dependency_wait: true
      agent: i2v-bytedance-seedance-v1-lite-reference-to-video
      api: []
      parameters:
        aspect_ratio: "16:9"
        resolution: "720p"
        duration: "5"
        camera_fixed: false
        seed: -1

src:
  structure.yaml:
    content: |-
      Project: Natural Motion Video Generation with ByteDance Seedance v1 Pro
      - High-quality image-to-video conversion
      - Camera motion control options
      - Reproducible generation with seed values
      - Multiple resolution support (720p/1080p)
    agent_selection_reason: Using i2v-bytedance-seedance-v1-pro for high-quality motion generation with camera control
    agent: t2i-flux-schnell
    dependency: []
  image-generation:
    base-image.png:
      content: |-
        A peaceful garden scene with blooming flowers
      agent: t2i-flux-schnell
      dependency: []
      api: []
  video-generation:
    garden-motion.mp4:
      content: |-
        Gentle breeze moving through the flowers, petals swaying naturally
        Butterflies flying between the flowers
        Soft sunlight filtering through leaves
      dependency:
        - src/image-generation/base-image.png
      dependency_wait: true
      agent: i2v-bytedance-seedance-v1-pro
      api: []
      parameters:
        resolution: "1080p"
        camera_fixed: false
        seed: -1

src:
  structure.yaml:
    content: |-
      Project: Talking Avatar with OmniHuman
      - Combine a portrait still with voice-over audio
      - Generate expressive lip-sync aligned to speech
    agent_selection_reason: Using i2v-bytedance-omnihuman-v1.5 for high-fidelity facial motion and lip-sync.
    agent: claude-3-5-sonnet-20241022
    dependency: []
  image-generation:
    avatar.png:
      content: |-
        A friendly presenter facing the camera with soft studio lighting.
      agent: t2i-flux-schnell
      dependency: []
      api: []
  audio:
    voice-line.mp3:
      content: |-
        "Welcome to our product showcase. Let me guide you through the highlights."
      agent: t2s-gpt-4o-mini-tts
      dependency: []
      api: []
  video-generation:
    talking-avatar.mp4:
      content: |-
        Synchronize the speech with natural facial expressions and subtle head movements.
      dependency:
        - src/image-generation/avatar.png
        - src/audio/voice-line.mp3
      dependency_wait: true
      agent: i2v-bytedance-omnihuman-v1.5
      api: []
      parameters:
        audio_url: src/audio/voice-line.mp3

src:
  structure.yaml:
    content: |-
      Project: High-Quality Video Generation with MiniMax Hailuo-02 Pro
      - Advanced image-to-video conversion with prompt optimization
      - Natural and sophisticated video generation
      - Enhanced detail and motion quality
    agent_selection_reason: Using i2v-minimax-hailuo-02-pro for enhanced video quality with prompt optimization
    agent: t2i-flux-schnell
    dependency: []
  image-generation:
    source-image.png:
      content: |-
        A vibrant cityscape at golden hour with bustling street life
      agent: t2i-flux-schnell
      dependency: []
      api: []
  video-generation:
    city-animation.mp4:
      content: |-
        Gentle movement of people walking through the streets
        Soft lighting changes as the golden hour progresses
        Natural urban atmosphere with subtle life and energy
        Camera slowly pans across the vibrant cityscape
      dependency:
        - src/image-generation/source-image.png
      dependency_wait: true
      agent: i2v-minimax-hailuo-02-pro
      api: []
      parameters:
        prompt_optimizer: true

src:
  structure.yaml:
    content: |-
      Project: Cinematic WAN 2.5 Preview shot
      - Generate a 10 second tracking shot from a still hero image
      - Maintain character fidelity while adding dynamic camera motion
      - Use 1080p output for presentation footage
    agent_selection_reason: Selecting i2v-wan-25-preview to leverage WAN 2.5 Preview's long clip support and smooth motion quality.
    agent: claude-3-5-sonnet-20241022
    dependency: []
  image-generation:
    hero-frame.png:
      content: |-
        A heroic warrior standing on a cliff at sunset, cape flowing in the wind.
      agent: t2i-flux-pro
      dependency: []
      api: []
  video-generation:
    cinematic-shot.mp4:
      content: |-
        Create a 10-second 1080p clip with subtle camera push-in and atmospheric motion around the character.
      dependency:
        - src/image-generation/hero-frame.png
      dependency_wait: true
      agent: i2v-wan-25-preview
      api: []
      parameters:
        duration: "5"
        resolution: "1080p"
        aspect_ratio: "16:9"

src:
  structure.yaml:
    content: |-
      Project: Speech to Avatar Video Project
      - 音声からアバター動画を生成
      - 30種類のアバターから選択可能
      - 自然な口パクとジェスチャー
    agent_selection_reason: Project structure definition
    dependency: []
    agent: claude-3-5-sonnet-20241022
    api: []
  audio:
    sample_speech.mp3:
      content: "こんにちは！私はVEED Avatarsを使用したアバターです。音声に合わせて自然に話すことができます。"
      agent_selection_reason: 音声ファイルの生成
      dependency: []
      agent: t2s-gpt-4o-mini-tts
      api: []
  video:
    avatar_video.mp4:
      content: |-
        音声からアバター動画を生成
        - Emily（縦向き・基本）アバターを使用
        - 音声に合わせた自然な口パクとジェスチャー
      agent_selection_reason: s2v-veed-avatarsは音声からリアルなアバター動画を生成でき、30種類のアバターから選択可能
      dependency:
        - src/audio/sample_speech.mp3
      dependency_wait: true
      agent: s2v-veed-avatars
      parameters:
        avatar_id: emily_vertical_primary
      api: []

src:
  structure.yaml:
    content: KAMUI Anime Avatar 8-Second Conversation Test
    agent_selection_reason: プロジェクト構造定義とKAMUI日本語アバター機能のテスト
    dependency: []
    agent: claude-3-5-sonnet-20241022
    api: []
  audio:
    japanese_greeting.mp3:
      content: みなさん、こんにちは！新しいアニメアバターが登場しました！
      agent_selection_reason: |-
        t2s-minimax-speech-02-turboは日本語の自然な発音と抑揚に対応しており、
        親しみやすい挨拶メッセージの生成に適しています。
      dependency: []
      agent: t2s-minimax-speech-02-turbo
      parameters:
        voice_setting:
          voice_id: Decent_Boy
          speed: 1.4
          emotion: neutral
        language_boost: Japanese
        audio_setting:
          format: mp3
          sample_rate: 44100
      api: []
    japanese_response.mp3:
      content: わあ、すごい！どんなことができるんですか？
      agent_selection_reason: |-
        t2s-minimax-speech-02-turboのDeep_Voice_Manを使用して男性の声で応答を生成。
        対話の自然な流れを作るため、異なる音声IDを使用します。
      dependency: []
      agent: t2s-minimax-speech-02-turbo
      parameters:
        voice_setting:
          voice_id: Lively_Girl
          speed: 1.2
          emotion: neutral
        language_boost: Japanese
        audio_setting:
          format: mp3
          sample_rate: 44100
      api: []
  image:
    conversation_scene.jpg:
      content: Animation Taste, Animation TasteAnimation Taste, Two realistic forest animals having an animated conversation in a beautiful woodland setting. A wise old brown bear sitting upright with expressive eyes and gentle facial expression, appearing to be speaking earnestly to a clever red fox who is sitting attentively with tilted head and bright, intelligent eyes showing curiosity. The bear's mouth is slightly open as if mid-sentence, with natural body language showing engagement. The fox has perked ears and an alert, listening posture. Surrounding them is a lush forest with dappled sunlight filtering through tall trees, creating a warm, magical atmosphere. Moss-covered rocks and wildflowers scattered around. Photorealistic style with incredible detail in fur texture, natural lighting, and authentic woodland environment. The scene captures the essence of two friends sharing stories in nature.

      agent_selection_reason: t2i-flux-proを使用してアニメテイストの男女キャラクターを生成。FLUX
      dependency: []
      agent: t2i-imagen4
      api: []
  video:
    kamui_japanese_avatar_conversation.mp4:
      content: |-
        Two Japanese professionals having a friendly conversation about the new KAMUI grimoire update in a modern office setting. 
        Natural dialogue flow with expressive gestures and eye contact, showcasing the advanced Japanese language avatar capabilities.
      agent_selection_reason: s2v-ai-avatar-multiはデュアル音声対応でWise_WomanとDeep_Voice_Manの2つの声を使った自然な日本語対話動画を生成
      dependency:
        - src/audio/japanese_greeting.mp3
        - src/audio/japanese_response.mp3
        - src/image/conversation_scene.jpg
      dependency_wait: true
      agent: s2v-ai-avatar-multi
      parameters:
        num_frames: 192
        turbo: true
        seed: 12345
      api: []

src:
  storyboard.yaml:
    content: |-
      Project: Futuristic City Reveal with Cinematic Lighting
      - From text prompt to Wan 2.5 Preview video generation
      - Lean camera move that circles around the hero subject
      - Optional background score supplied via audio_url
      - Use 1080p resolution and 10 second duration for presentation reels
      - Deliver both vertical and landscape cuts for social media
      - Include fallback prompts for negative details (noise, low quality)
    agent_selection_reason: Gemini 2.5 Flash organizes requirements and prepares prompts for multi-agent video generation.
    dependency: []
    agent: gemini-2.5-flash-preview-04-17
    api: []
video-generation:
  wan25_preview_demo.mp4:
    content: |-
      Cinematic reveal of a cyberpunk city skyline at sunset.
      The camera begins with a wide aerial shot, then glides around a lone android hero standing on a balcony.
      Neon reflections ripple across glass towers while drones pass overhead.
      Include atmospheric haze, volumetric light shafts, and dynamic clouds.
      Keep movements smooth and emphasize premium cinematic lighting.
    agent_selection_reason: Wan 2.5 Preview handles 1080p output, long motion, and nuanced lighting in 10 second clips.
    dependency: []
    agent: t2v-wan-25-preview
    api: []
    parameters:
      duration: "5"
      aspect_ratio: "16:9"
      resolution: "1080p"
      enable_prompt_expansion: true
      negative_prompt: "low resolution, muted colors, jitter, artifacts"

src:
  video:
    magical_forest.mp4:
      content: |-
        A magical forest at dawn with glowing fireflies dancing between ancient trees. 
        Soft mist rolls through the undergrowth as golden sunlight filters through the canopy. 
        Ethereal and cinematic atmosphere with smooth camera movement.
      dependency: []
      agent: "t2v-kling-2.1"
      parameters:
        duration: 10                               # 5または10秒を指定
        aspect_ratio: "16:9"                       # 16:9, 9:16, 1:1から選択
        style: "cinematic fantasy"                 # スタイル指定（プロンプトに追加）
        negative_prompt: "low quality, blurry, static camera"  # 除外したい要素
        cfg_scale: 0.7                            # プロンプトへの忠実度（0.0-1.0）
        seed: 123456                              # 再現性のためのシード値（オプション）
      api: []
      
    # 最小構成の例
    simple_scene.mp4:
      content: "A serene lake at sunset with rippling water"
      dependency: []
      agent: "t2v-kling-2.1"
      api: []  # parametersを省略するとデフォルト値が使用されます

src:
  video:
    futuristic_city.mp4:
      content: |-
        A futuristic city skyline at sunset with flying cars weaving between holographic billboards.
        - Neon reflections on glass buildings
        - Dynamic aerial traffic
        - Warm cinematic lighting
      dependency: []
      agent: "t2v-kling-2.5-turbo-pro"
      parameters:
        duration: "5"
        aspect_ratio: "16:9"
        cfg_scale: 0.6
        seed: 12345
      api: []

src:
  structure.yaml:
    content: |-
      Project: Speech to Avatar Video Project
      - 音声からアバター動画を生成
      - 30種類のアバターから選択可能
      - 自然な口パクとジェスチャー
    agent_selection_reason: Project structure definition
    dependency: []
    agent: claude-3-5-sonnet-20241022
    api: []
  audio:
    sample_speech.mp3:
      content: "こんにちは！私はVEED Avatarsを使用したアバターです。音声に合わせて自然に話すことができます。"
      agent_selection_reason: 音声ファイルの生成
      dependency: []
      agent: t2s-gpt-4o-mini-tts
      api: []
  video:
    avatar_video.mp4:
      content: |-
        音声からアバター動画を生成
        - Emily（縦向き・基本）アバターを使用
        - 音声に合わせた自然な口パクとジェスチャー
      agent_selection_reason: s2v-veed-avatarsは音声からリアルなアバター動画を生成でき、30種類のアバターから選択可能
      dependency:
        - src/audio/sample_speech.mp3
      dependency_wait: true
      agent: s2v-veed-avatars
      parameters:
        avatar_id: emily_vertical_primary
      api: []