Images

Synchronous image generation. For long-running or queue-based image jobs (Midjourney imagine → upscale workflows), use the async tasks endpoint instead.

Pick an image model#

Image parameters vary by model — gpt-image-2 takes size + quality, Gemini and Flux take aspect_ratio, Midjourney needs a multi-step workflow. The table below lets you filter; the per-model section that follows shows every parameter.

Request body (shared)#

Response#

Per-model parameters#

Each model exposes its own parameter set. Pass only what the table for that model lists — Orux AI rejects unknown fields with invalid_param so misconfiguration shows up at request time rather than as silently dropped behaviour.

Midjourney workflow: imagine → upscale / variation / reroll#

Midjourney is a multi-step queue. imagine returns a 4-image grid and a task_id; to obtain a single high-resolution image you must send a follow-up request with action=upscale (or variation, reroll) referencing the original task_id and an index (1–4).

┌──────────────────┐
│  POST imagine    │  → task_id A,  4-image grid (low-res)
└────────┬─────────┘
         │
         ├─→ POST { action: "upscale",   parent_task_id: A, index: 2 }  → high-res single image
         ├─→ POST { action: "variation", parent_task_id: A, index: 1 }  → new 4-image grid
         └─→ POST { action: "reroll",    parent_task_id: A }            → re-run same prompt

Reference images#

Several models (flux-kontext-pro/max, runway-gen4-image, midjourney-v7) accept a reference_image_url (or reference_images array). Pass an https URL to a publicly accessible JPG/PNG/WEBP. For Midjourney, the upscale / variation actions implicitly reference the parent task_id — no URL needed.

Examples#

curl https://orux.top/api/v1/images/generations \
  -H "Authorization: Bearer $ORUX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2",
    "prompt": "A cinematic shot of a fox in a neon-lit alley, rain",
    "size": "1024x1024",
    "n": 1,
    "response_format": "url"
  }'

curl https://orux.top/api/v1/images/generations \
  -H "Authorization: Bearer $ORUX_API_KEY" \
  -d '{
    "model":"gpt-image-2",
    "prompt":"Cinematic neon-lit Tokyo alley in the rain",
    "size":"1536x1024",
    "quality":"high",
    "n":1
  }'

curl https://orux.top/api/v1/images/generations \
  -H "Authorization: Bearer $ORUX_API_KEY" \
  -d '{
    "model":"flux-kontext-pro",
    "prompt":"a photo of the subject riding a bicycle, golden hour",
    "aspect_ratio":"3:4",
    "reference_image_url":"https://example.com/subject.jpg",
    "prompt_upsampling": true,
    "seed": 42
  }'