MAI Image 2.6

If you want a picture that looks like it came out of a product studio, MAI Image 2.6 is a good first try. Microsoft AI built it in-house, and it handles two jobs: drawing a new image from a written brief, and reworking a photo you give it with the help of up to five reference images. Each run hands back a single PNG.

What you can make

• Packshots and lifestyle scenes for a shop, lit like a real photo session.
• Article and slide graphics at 2K that stay sharp on a large screen.
• Very wide 4:1 headers for a website or newsletter, and tall 1:4 strips for side columns.
• Fixed-up versions of an existing photo: new background, cleaned surface, swapped prop, steered by your references.

Settings that matter

• Resolution: 1K or 2K. Iterate at 1K while the idea is still moving, switch to 2K for the version you keep.
• Aspect ratio: 13 fixed shapes between 4:1 and 1:4, plus auto. With auto the model reads your text and chooses a frame, so pick a fixed value whenever the slot on your page has a set size.
• Editing mode: once you attach references, the frame and size are decided by the model (roughly one megapixel). The ratio and resolution fields only steer generation from text.
• References: up to 5 images. Fewer, clearer references usually give a more predictable edit than five loosely related ones.
• Output: one image per run. For three variants, send three runs.

Starter prompt

Ceramic coffee mug on a pale oak table, soft window light from the left, shallow depth of field, steam rising, neutral beige backdrop, catalog photo, no text

Swap the mug and the table for your own product and surface, and keep the light direction so a whole series matches.

Questions

What will one MAI Image 2.6 picture cost me?

from $0.0934 per image, with 1K and 2K billed at different rates (both are in the table above and on the pricing page). You see the exact amount on the Generate button before anything runs, and if the job fails you pay nothing.

My edited photo came back in a different shape. Did I set something wrong?

No. In edit mode the model sizes the result from the photo you supplied and ignores the ratio field. If you need a strict 16:9, generate from text or crop afterwards.

Can I get several variants from one request?

No, each request returns one picture. Send separate runs, and if you work through the API, add a webhook_url to POST /v1/generate so results arrive as they finish (docs).

MAI Image 2.6 in your code

curl https://kadraris.com/v1/generate?wait=60 \
  -H "Authorization: Bearer $KADRARIS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "mai-image-2.6",
  "input": {
    "prompt": "Packshot of a white linen sneaker on a concrete block, hard midday sun, crisp shadow, 85mm",
    "aspect_ratio": "1:1",
    "resolution": "1K"
  }
}'

import os, time, requests

API = "https://kadraris.com/v1"
H = {"Authorization": f"Bearer {os.environ['KADRARIS_API_KEY']}"}

task = requests.post(f"{API}/generate", headers=H, json={
    "model": "mai-image-2.6",
    "input": {
        "prompt": "Packshot of a white linen sneaker on a concrete block, hard midday sun, crisp shadow, 85mm",
        "aspect_ratio": "1:1",
        "resolution": "1K"
    }
}).json()
while task["status"] in ("queued", "running"):
    time.sleep(3)
    task = requests.get(f"{API}/tasks/{task['id']}", headers=H).json()

print(task["status"], task["cost"], [o["url"] for o in task["output"]])

const API = "https://kadraris.com/v1";
const headers = { Authorization: `Bearer ${process.env.KADRARIS_API_KEY}`, "Content-Type": "application/json" };

let task = await (await fetch(`${API}/generate`, {
  method: "POST",
  headers,
  body: JSON.stringify({
    "model": "mai-image-2.6",
    "input": {
      "prompt": "Packshot of a white linen sneaker on a concrete block, hard midday sun, crisp shadow, 85mm",
      "aspect_ratio": "1:1",
      "resolution": "1K"
    }
  }),
})).json();

while (task.status === "queued" || task.status === "running") {
  await new Promise((r) => setTimeout(r, 3000));
  task = await (await fetch(`${API}/tasks/${task.id}`, { headers })).json();
}
console.log(task.status, task.cost, task.output.map((o) => o.url));

Input fields

Fetch this list as JSON with GET /v1/models/mai-image-2.6.