AI Model Wiki

    Skip to main content
    35 min
    Lesson 3 of 5

    Upscaling, Video & Advanced Features

    From the course Krea AI: Real-Time Image Generation

    Explore Krea's advanced features for professional-grade output and video content creation.

    Image Upscaling

    AI Upscaler

    • Enhance image resolution up to 4x without quality loss
    • AI adds detail and sharpness during upscaling
    • Works on AI-generated and real photographs
    • Multiple upscaling models for different content types

    Upscaling Use Cases

    • Prepare AI images for print production
    • Enhance old or low-resolution photos
    • Scale up thumbnails for larger display
    • Improve detail in generated textures and patterns

    Settings

    • Scale factor: 2x or 4x resolution increase
    • Enhancement level: How much AI-generated detail to add
    • Face enhancement: Special handling for facial features
    • Noise reduction: Clean up artifacts during upscaling

    Video Generation

    Text-to-Video

    • Enter a prompt describing the desired video
    • Krea generates short video clips (3-5 seconds)
    • Control motion type: camera movement, subject motion
    • Specify style: cinematic, animation, abstract

    Image-to-Video

    • Upload a static image as the starting frame
    • Describe the desired motion and animation
    • AI animates the image naturally
    • Maintain the visual style of the source image

    Video Settings

    • Duration and frame rate options
    • Motion intensity controls
    • Camera movement preferences
    • Style consistency across frames

    Advanced Generation

    ControlNet Integration

    Guide generation with structural controls:

    • Depth maps: Control spatial arrangement
    • Edge detection: Maintain outlines from reference images
    • Pose estimation: Match human poses from reference
    • Segmentation maps: Define regions for different content

    Batch Generation

    • Generate multiple images with variations
    • Systematic prompt exploration
    • Save successful configurations
    • Build visual asset libraries efficiently

    Best Practice: When using ControlNet depth maps, create your depth map from a simple 3D render in Blender rather than estimating depth from a photo. Precise depth control produces significantly more accurate spatial arrangements, especially for product visualization.

    Common Pitfall: Batch generation burns through credits quickly. Test your best prompt on a single generation first, then batch-generate variations only after you are satisfied with the style and composition. Batch mode is for volume, not experimentation.

    ControlNet workflow for consistent character design:
    1. Create a reference pose in a stick-figure sketch
    2. Upload as the pose control input
    3. Set ControlNet type: OpenPose
    4. Write character description prompt
    5. Generate multiple style variations using the same pose
    

    Key Takeaways

    • •AI upscaling enhances resolution up to 4x with added detail for print-ready output
    • •Video generation creates short clips from text prompts or animated from static images
    • •ControlNet integration provides structural guidance through depth maps, edges, and poses
    • •Batch generation enables systematic visual exploration and efficient asset library creation
    0 of 5 lessons marked complete

    Progress is kept in this browser only.