ArchDiffusion v4.6

ArchDiffusion v4.6 is the engine that powers the mnml studio today. It renders on a new model at a 2048-pixel long edge, and it treats a render as a photograph of the finished building, made for an architecture magazine, rather than as a render to be polished. Your image is kept exactly: the design, the camera and the framing.

The prompt this endpoint composes is the studio's prompt, byte for byte, for the same settings. The same settings on this API and in the studio send the model the same brief.

Endpoint

POST https://api.mnmlai.dev/v1/archDiffusion-v46

Request

Send a POST request with multipart/form-data. Do not set the Content-Type header by hand; let your HTTP client set it, with its boundary. The API answers at once with a request ID; poll the status endpoint for the image.

Common Parameters

Every parameter below is checked against the expert you name. Values match case-insensitively, by value or by label, and spaces, underscores and hyphens are treated alike, so living_room and Living Room both work. A parameter that belongs to another expert is ignored. Defaults marked auto are left to the engine; unset means nothing is sent unless you pick a value.

Response

parameters lists the settings the brief was composed from, including the defaults you did not send. When your words moved a setting, parameters.from_prompt says from what to what.

{
  "status": "success",
  "id": "7c1f3e0a-2b9d-4c61-9d7e-0f3a8b2c4d11",
  "prompt": "the museum at night, lit from within",
  "expert_name": "exterior",
  "parameters": {
    "prompt_mode": "render",
    "render_style": "auto",
    "camera_angle": "auto",
    "camera_direction": "front",
    "time_of_day": "night",
    "site_context": "auto",
    "motion": "off",
    "aspect_ratio": "auto",
    "reference_image_count": 0,
    "from_prompt": {
      "time_of_day": { "from": "golden_hour", "to": "night" }
    }
  },
  "credits": 97
}

Checking Processing Status

GET https://api.mnmlai.dev/v1/status/{id}

Code Examples

curl -X POST https://api.mnmlai.dev/v1/archDiffusion-v46 \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "[email protected]" \
  -F "expert_name=exterior" \
  -F "render_style=auto" \
  -F "time_of_day=blue_hour" \
  -F "people=few" \
  -F "prompt=a timber pavilion by a lake"

curl -X POST https://api.mnmlai.dev/v1/archDiffusion-v46 \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "[email protected]" \
  -F "prompt_mode=edit" \
  -F "prompt=replace the facade cladding with the brick in image 2" \
  -F "[email protected]" \
  -F "reference_mode_1=material"

curl -X POST https://api.mnmlai.dev/v1/archDiffusion-v46 \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "expert_name=text-to-render" \
  -F "prompt=a concrete chapel on a sea cliff" \
  -F "time_of_day=dusk" \
  -F "aspect_ratio=16:9"

const fs = require('fs');

const form = new FormData();
form.append('image', new Blob([fs.readFileSync('room.png')]), 'room.png');
form.append('expert_name', 'interior');
form.append('room_type', 'living room');
form.append('room_style', 'Japandi');
form.append('interior_accessories', 'on');
form.append('ambience', 'golden_hour');

const res = await fetch('https://api.mnmlai.dev/v1/archDiffusion-v46', {
  method: 'POST',
  headers: { Authorization: 'Bearer YOUR_API_KEY' },
  body: form,
});
const { id } = await res.json();

Migrating from v4.5

Error Handling

// 400 — a value this expert does not offer
{
  "status": "error",
  "code": "INVALID_PARAMETER",
  "message": "Unsupported motion value for the \"exterior\" expert. Allowed values: off, long_exposure",
  "details": { "field": "motion", "received": "subtle", "allowed": ["off", "long_exposure"] }
}

// 400 — an edit with no instruction
{ "status": "error", "code": "MISSING_PROMPT", "message": "prompt_mode=edit needs a prompt: the instruction to apply to the image" }

// 400 — no image, on an expert that needs one
{ "status": "error", "code": "MISSING_IMAGE", "message": "Image file is required" }

// 400 — not enough credits
{
  "status": "error",
  "code": "NO_CREDITS",
  "message": "You do not have enough credits to use this feature. This API requires 3 credits.",
  "details": { "credits": 2, "required": 3 }
}

// Also: IMAGE_TOO_SMALL, IMAGE_TOO_LARGE, INVALID_IMAGE_TYPE (for image or a reference),
// PROMPT_TOO_LONG, INVALID_REQUEST_BODY.