Skip to main content

Video Generation

Getting Started​

Start from GUI​

Menu: AI > Generate > Generate Video, or refer to the getting started section of 3D Model Generation.

Start from CLI​

Use the generate_video tool to generate video. After generation, the video is automatically imported as VideoClip and ready to use in your scene.

Input Modes:

ModeParametersDescription
Text-to-Videoprompt + mode: "text_to_video"Pure text description
Image-to-Videoprompt + image_path (mode defaults to reference_image)Reference image + text guidance
Scene-to-VideoScene screenshot as image_path + promptGenerate transition/promo video from current scene screenshot

⚠️ The tool default mode is reference_image; for pure text-to-video, explicitly pass mode: "text_to_video". For scene-to-video, ratio must match the screenshot aspect ratio.

⚠️ The prompt parameter must be in English; Chinese prompts cannot generate correct results.

Models​

Huoshan SeeDance (Only Model)​

  • generator_id: huoshan-seedance
  • Use Cases: Concept video, motion graphics, visual effects, transition video, promo video, cutscene animation
  • Output: Video file automatically imported as VideoClip, saved to Assets/Video/
  • Key Parameters:
    • prompt (required, must be English): Video description
    • mode: "reference_image" (default) / "text_to_video"; automatically switches to reference_image when image_path is provided
    • image_path: Reference image path (image-to-video mode)
    • resolution: "480p" (854×480) / "720p" (1280×720, default)
    • ratio: "16:9" (default) / "9:16" / "1:1"
    • duration: Video duration (integer 3–15 seconds, default 12)
    • return_last_frame: Whether to return the last frame as a preview image (default true)

Optimization​

Prompt Optimization (⚠️ Must be English)​

  • Describe dynamic changes: "a majestic dragon flying over a medieval castle, cinematic lighting, 4K quality" > "a dragon"
  • Specify camera movement: "slow zoom into..." / "panning shot of..." / "aerial view of..."
  • Include visual details: "cinematic lighting" / "dramatic shadows" / "volumetric fog"

Scene-to-Video Workflow​

  • You can use a screenshot of the current Unity scene as a reference image to generate transition/promo videos
  • Suitable for generating cutscene animations and promotional assets

Resolution & Duration​

  • 480p: Quick preview, smaller file size
  • 720p: Recommended output quality
  • 3–5 seconds: Transition/short video effects
  • 10–15 seconds: Promo/concept video

Parallel Generation​

  • Multiple video generation tasks can be submitted simultaneously

Notes​

  • ⚠️ Prompts must be in English, Chinese prompts will produce incorrect results; if the user describes in Chinese, translate to English first
  • ⚠️ Generation takes 30–120 seconds
  • ⚠️ Video duration is limited (3–15 seconds), not suitable for long videos
  • ⚠️ Maximum resolution is 720p, for higher resolution use other tool chains
  • ⚠️ VideoPlayer configuration: After generation, check VideoPlayer's source, clip, loop, play_on_awake settings
  • ⚠️ Output path: Videos are saved to Assets/Video/ by default (different from other types, not Assets/TJGenerators/History/)
  • ⚠️ Domain Reload: Do not write .cs files to disk during generation