name: Stability AI Vocabulary
description: >-
  Vocabulary for Stability AI generative AI models and APIs covering image
  generation, image editing, upscaling, video generation, and 3D asset
  creation including Stable Diffusion and related concepts.
created: '2026-05-02'
modified: '2026-05-02'
terms:
  - term: Stable Diffusion
    definition: >-
      A deep learning text-to-image generation model developed by Stability AI.
      Uses a latent diffusion process where noise is iteratively removed from
      a latent representation to produce high-quality images from text prompts.

  - term: Text-to-Image
    definition: >-
      The process of generating images from natural language text descriptions
      (prompts). Stability AI's core capability through models like Stable
      Image Core, Stable Diffusion 3, and Stable Image Ultra.

  - term: Image-to-Image
    definition: >-
      Generating a new image guided by an input reference image and a text
      prompt. The strength parameter controls how much the output diverges from
      the input image.

  - term: Prompt
    definition: >-
      A text description that guides the image generation model. Strong prompts
      clearly describe subjects, styles, lighting, and composition. Maximum
      10,000 characters in the Stability AI API.

  - term: Negative Prompt
    definition: >-
      Text describing elements to exclude from the generated image. Used to
      steer the model away from unwanted content, artifacts, or styles.

  - term: Seed
    definition: >-
      An integer value that initializes the random number generator used during
      image generation. Using the same seed with identical parameters produces
      similar results, enabling reproducibility.

  - term: Aspect Ratio
    definition: >-
      The proportional relationship between width and height of the generated
      image. Supported values include 1:1, 16:9, 9:16, 4:5, 5:4, 2:3, 3:2,
      21:9, and 9:21.

  - term: Output Format
    definition: >-
      The file format of the generated image. Stability AI APIs support JPEG,
      PNG, and WebP output formats. PNG preserves transparency for background
      removal operations.

  - term: Inpainting
    definition: >-
      The process of filling in a masked region of an image with AI-generated
      content. A mask image identifies the area to modify; a text prompt guides
      what appears in the filled region.

  - term: Outpainting
    definition: >-
      The process of extending an image beyond its current boundaries by
      generating new content that seamlessly continues the existing scene.

  - term: ControlNet
    definition: >-
      A neural network architecture for controlling image generation using
      additional conditioning inputs such as sketches, depth maps, or
      segmentation masks. Used in Stability AI's Stable Image Control API.

  - term: Stable Image Ultra
    definition: >-
      Stability AI's highest quality image generation model tier, offering
      state-of-the-art photorealistic and artistic image generation with
      exceptional detail and composition.

  - term: Stable Image Core
    definition: >-
      Stability AI's fast and affordable image generation model tier, optimized
      for quick, high-quality generation suitable for most use cases.

  - term: Stable Diffusion 3
    definition: >-
      The third generation of Stability AI's Stable Diffusion model family,
      offering improved text rendering, multi-subject composition, and overall
      image quality. Variants include SD3, SD3 Turbo, SD3.5 Large, and SD3.5
      Medium.

  - term: Stable Video Diffusion
    definition: >-
      Stability AI's image-to-video model that generates short animated video
      clips from a single input image, producing realistic motion from static
      images.

  - term: Stable Fast 3D
    definition: >-
      Stability AI's rapid 3D asset generation model that creates textured 3D
      meshes from single input images, suitable for games, AR, and VR.

  - term: Finish Reason
    definition: >-
      The reason why a generation request completed. SUCCESS indicates a
      successful generation; CONTENT_FILTERED indicates the output was blocked
      by content moderation policies.

  - term: Content Moderation
    definition: >-
      Automatic filtering applied to generation requests and outputs to prevent
      the creation of harmful, offensive, or policy-violating content.

  - term: Bearer Token
    definition: >-
      Authentication method used by Stability AI APIs. The API key is passed
      as a Bearer token in the Authorization header of each request.

  - term: Multipart Form Data
    definition: >-
      The request encoding format used by Stability AI APIs for image generation
      and editing. Allows uploading both text parameters and binary image files
      in a single request.