# Built-in image-generation platforms.
#
# To add a new platform: append an entry under `platforms:` with a unique `id`,
# a human label, a one-line description, and the platform's syntax hints.
# Restart the server (or rebuild) to pick up the change.
#
# Currency policy (1.16.0): entries touched by a fact-verification pass carry
# `verifiedAt` (date facts were checked against current docs). Entries without
# a date have not been re-verified recently — treat their hints as suspect.

category:
  id: image
  label: Image
  description: Image generation
  defaultPlatform: midjourney
  defaultMode: detailed

platforms:
  - id: midjourney
    label: Midjourney (V8.x)
    description: Artistic, stylized imagery; current default is V8.2 — Omni Reference replaced by the Edit Model in V8.x
    verifiedAt: "2026-10-05"
    source: docs.midjourney.com
    syntaxHints:
      - "--ar W:H (aspect ratio)"
      - "--v 8.2 (current default; V7 introduced --draft and --oref)"
      - "--style raw"
      - "--stylize / --s 0-1000 (default 100)"
      - "--chaos / --c 0-100"
      - "--quality / --q (V7+: 1, 2, 4)"
      - "--hd / --sd (V8.1: 2048px vs 1024px output)"
      - "--sref <url|code> style reference, --sw weight"
      - "--oref <url> Omni Reference (V7), replaced by Edit Model in V8.x"
      - "--edit (V8.x Edit Model: modify with written instructions + up to 4 reference images)"
      - "--draft (V7: half GPU cost)"
      - "--p / --profile personalization"

  - id: dall-e
    label: OpenAI image (GPT Image 2 / DALL-E 3)
    description: Natural-language prompting; GPT Image 2 is OpenAI's current image model, DALL-E 3 still available in ChatGPT legacy
    verifiedAt: "2026-10-05"
    syntaxHints:
      - natural language, full sentences (no flags)
      - describe the whole scene conversationally
      - specify lighting, camera, style explicitly
      - state aspect ratio in words ("wide 16:9 landscape")
      - reliable in-image text when quoted

  - id: stable-diffusion
    label: Stable Diffusion
    description: Open source, highly customizable
    syntaxHints:
      - negative prompts
      - CFG scale
      - steps
      - samplers
      - LoRA
      - embeddings

  - id: flux
    label: Flux (2)
    description: Black Forest Labs — high detail, photorealistic, strong prompt adherence
    verifiedAt: "2026-10-05"
    syntaxHints:
      - natural language
      - high detail focus
      - guidance scale (typically 3-7; Flux tolerates low guidance)
      - multi-reference editing in Flux 2

  - id: ideogram
    label: Ideogram
    description: Best for text in images
    syntaxHints:
      - magic prompt
      - text rendering
      - typography

  - id: leonardo
    label: Leonardo AI
    description: Preset styles, game art
    syntaxHints:
      - preset styles
      - guidance scale
      - contrast
      - alchemy

  - id: firefly
    label: Adobe Firefly
    description: Commercial safe, natural
    syntaxHints:
      - natural language
      - style references
      - effects

  - id: grok-aurora
    label: Grok Aurora
    description: xAI, fast and creative
    syntaxHints:
      - natural language
      - creative interpretation
      - fast generation

  - id: imagen
    label: Google Imagen (4 / 4 Ultra)
    description: Google's current photorealistic image model (Imagen 4 line), via Gemini/Vertex
    verifiedAt: "2026-10-05"
    syntaxHints:
      - natural language
      - photorealistic
      - aspect ratios
      - no parameters in the prompt — controls live in the API call

  - id: nano-banana
    label: Nano Banana (Gemini 2.5 Flash Image)
    description: Natural-language scene direction; character consistency, multi-image edits, in-image text
    verifiedAt: "2026-10-05"
    syntaxHints:
      - describe the whole scene in natural language (direct the scene, don't list keywords)
      - photographic terminology to control the look (camera/lens, angle, depth of field)
      - explicit lighting (e.g. "three-point softbox setup", "golden-hour backlight")
      - editing — state what CHANGES and what STAYS the same (preserves subject identity)
      - combine multiple reference images for character consistency / product placement
      - reliable in-image text rendering
      - specify aspect ratio

  - id: recraft
    label: Recraft
    description: Vector design, brand assets
    syntaxHints:
      - style selection
      - vector output
      - brand colors
      - SVG export

  - id: higgsfield
    label: Higgsfield (multi-model platform)
    description: Multi-model creative platform via MCP / CLI / Skills. Routes one natural-language prompt to Soul 2.0, Soul Cinema, Soul Cast, Flux 2, Seedream 5, Nano Banana Pro, or GPT Image 2. Soul ID for face-faithful character reuse. No API key — auth via Higgsfield account.
    verifiedAt: "2026-10-05"
    syntaxHints:
      - "models: soul-2.0 | soul-cinema | soul-cast | flux-2 | seedream-5 | nano-banana-pro | gpt-image-2"
      - long-form natural-language prose (composition + lighting + textures + mood)
      - Soul ID for face-faithful identity reuse
      - multi-reference compositing for edits
      - output up to 4K
      - "modes: Marketing Studio, Fashion Factory, Photodump Studio"