midjourney-v7

Quick example

curl https://reapi.ai/api/v1/images/generations \
  -H "Authorization: Bearer rk_live_xxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "mj-v7",
    "prompt": "a serene Japanese garden with cherry blossoms --ar 16:9 --s 500",
    "model_params": { "speed": "fast" }
  }'

import requests

resp = requests.post(
    "https://reapi.ai/api/v1/images/generations",
    headers={
        "Authorization": "Bearer rk_live_xxx",
        "Content-Type": "application/json",
    },
    json={
        "model": "mj-v7",
        "prompt": "a serene Japanese garden with cherry blossoms --ar 16:9 --s 500",
        "model_params": {"speed": "fast"},
    },
    timeout=30,
)
task_id = resp.json()["id"]

const resp = await fetch('https://reapi.ai/api/v1/images/generations', {
  method: 'POST',
  headers: {
    Authorization: 'Bearer rk_live_xxx',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    model: 'mj-v7',
    prompt: 'a serene Japanese garden with cherry blossoms --ar 16:9 --s 500',
    model_params: { speed: 'fast' },
  }),
});
const { id } = await resp.json();

body := strings.NewReader(`{
  "model": "mj-v7",
  "prompt": "a serene Japanese garden with cherry blossoms --ar 16:9 --s 500",
  "model_params": { "speed": "fast" }
}`)
req, _ := http.NewRequest("POST", "https://reapi.ai/api/v1/images/generations", body)
req.Header.Set("Authorization", "Bearer rk_live_xxx")
req.Header.Set("Content-Type", "application/json")
resp, _ := http.DefaultClient.Do(req)

Generation is asynchronous. The call returns a task_id; poll GET /api/v1/tasks/:task_id until status is completed, then read the image URLs. Each request returns 1–4 images (content review may filter some) and is billed once per request.

Endpoint

POST https://reapi.ai/api/v1/images/generations
GET  https://reapi.ai/api/v1/tasks/:task_id

Models

The family is one base generator plus eleven edit operations. Each is a separate model id on the same endpoint.

The eight parent-referencing edits take the reAPI task_id of a completed Midjourney V7 generation in model_params.task_id, plus model_params.image_number (0–3) to pick which of the four grid images to act on. reAPI resolves your task id to the upstream reference automatically.

Base generation — mj-v7

Native MJ parameters are written inside the prompt, not as JSON fields:

--v / --version / --niji are ignored (the version is locked to V7). Speed is set via model_params.speed, not a prompt flag.

Edit operations

All edits share the endpoint and the async task_id lifecycle. speed (where present) is fast / turbo (default fast).

Parent-referencing edits

Common model_params: task_id (required, the reAPI task id of a completed Midjourney V7 generation), image_number (0–3, default 0).

mask is one of two shapes (mutually exclusive):

{ "mask": { "areas": [ { "width": 200, "height": 200, "points": [50,50,50,250,250,250,250,50] } ] } }

{ "mask": { "url": "https://example.com/mask.png" } }

points is a flat [x1,y1,x2,y2,…] closed polygon; for the URL form, white = repaint region, black = preserve. canvas is width / height (px); img_pos is width / height / x / y (placement of the source image on the canvas).

Example — upscale image 0 of a prior generation:

{
  "model": "mj-v7-upscale",
  "model_params": { "task_id": "task_xxx", "image_number": 0, "type": "standard" }
}

Direct-image edits

These take the source image in image_urls (exactly one public http(s) URL) and do not reference a prior task.

Pricing

Midjourney V7 bills per request (four images), not per image. Cost scales with:

• Speed tier — draft < fast < turbo.
• Operation — base generation and lightweight edits (remove-bg, pan, outpaint, enhance) cost less than heavy edits (upscale, variation, remix, inpaint, canvas edit, retexture, upload-paint).
• --oref — Omni Reference in the prompt doubles the cost.

Integer credits, 1 credit = $0.001:

credits = ceil(request_price_usd × 1000)

where request_price_usd is the (model, speed) rate, doubled when the prompt carries --oref. See the model page for current rates.

Output

Poll GET /api/v1/tasks/:task_id. On completed:

{
  "id": "task_xxx",
  "status": "completed",
  "output": { "image_urls": ["https://cdn.reapi.ai/.../0.png", "https://cdn.reapi.ai/.../1.png"] }
}

output.image_urls holds 1–4 URLs (content review may filter some of the four). Save them promptly.

Errors

Customer-facing errors use the standard envelope { error: { code, message, request_id } }. Common cases:

See the full catalog at /docs/api/errors.

Tips

• A single reference image with no text is rejected — add a short description.
• Use --oref for subject consistency and --sref for style transfer; they combine. --oref doubles the bill, so reserve it for shots that need it.
• Generate with mj-v7 first, then chain mj-v7-upscale / mj-v7-variation / mj-v7-inpaint off the returned task_id — no re-upload needed.
• mj-v7-enhance only accepts a parent created with speed: "draft".
• Keep the version implicit — --v / --niji are ignored on V7.

Related

• Model page & live playground
• API quickstart
• Error codes