grok-imagine-image-2-0

Quick example

curl https://reapi.ai/api/v1/images/generations \
  -H "Authorization: Bearer rk_live_xxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "grok-imagine-image-2-0",
    "prompt": "An editorial product poster for a ceramic pour-over kettle, bold headline type, soft studio light, generous negative space",
    "n": 4,
    "size": "3:2"
  }'

import requests

resp = requests.post(
    "https://reapi.ai/api/v1/images/generations",
    headers={
        "Authorization": "Bearer rk_live_xxx",
        "Content-Type": "application/json",
    },
    json={
        "model": "grok-imagine-image-2-0",
        "prompt": "An editorial product poster for a ceramic pour-over kettle, bold headline type, soft studio light, generous negative space",
        "n": 4,
        "size": "3:2",
    },
    timeout=30,
)
print(resp.json())

const r = 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: "grok-imagine-image-2-0",
    prompt:
      "An editorial product poster for a ceramic pour-over kettle, bold headline type, soft studio light, generous negative space",
    n: 4,
    size: "3:2",
  }),
});
console.log(await r.json());

package main

import (
    "bytes"
    "encoding/json"
    "fmt"
    "io"
    "net/http"
)

func main() {
    body, _ := json.Marshal(map[string]any{
        "model":  "grok-imagine-image-2-0",
        "prompt": "An editorial product poster for a ceramic pour-over kettle, bold headline type, soft studio light, generous negative space",
        "n":      4,
        "size":   "3:2",
    })
    req, _ := http.NewRequest("POST", "https://reapi.ai/api/v1/images/generations", bytes.NewReader(body))
    req.Header.Set("Authorization", "Bearer rk_live_xxx")
    req.Header.Set("Content-Type", "application/json")

    resp, _ := http.DefaultClient.Do(req)
    defer resp.Body.Close()
    out, _ := io.ReadAll(resp.Body)
    fmt.Println(string(out))
}

The response carries a task_id. Poll GET /api/v1/tasks/{task_id} until status is completed, then read result.images[].url.

Channels

This model is served by two independent channels with different parameter surfaces. Pick one with the model field:

They are not interchangeable: a size value is rejected by the official channel and an aspect_ratio value is rejected by the standard one. The resolution field exists on both but means different things.

Request body

Fields below apply to the standard channel (grok-imagine-image-2-0).

size values

Seven aspect ratios:

Five pixel aliases are also accepted: 1024x1024 (1:1), 1024x1792 (2:3), 1792x1024 (3:2), 720x1280 (9:16), 1280x720 (16:9). Ratios not on this list — 1:2, 2:1, 4:5, auto — return 400 INVALID_REQUEST.

Official channel

Send "model": "grok-imagine-image-2.0-official" to use the vendor's own parameter surface. It exposes wider framing control, a real pixel tier, a quality setting, and image-to-image.

aspect_ratio values

1:1, 3:4, 4:3, 9:16, 16:9, 2:3, 3:2, 9:19.5, 19.5:9, 9:20, 20:9, 1:2, 2:1, auto.

auto lets the model pick the framing that suits the prompt. Values outside this list return 400 INVALID_REQUEST. Note the standard channel's size values are not accepted here, and these values are not accepted there.

Text-to-image on the official channel

curl https://reapi.ai/api/v1/images/generations \
  -H "Authorization: Bearer rk_live_xxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "grok-imagine-image-2.0-official",
    "prompt": "An editorial product poster for a ceramic pour-over kettle, bold headline type, soft studio light, generous negative space",
    "n": 1,
    "aspect_ratio": "16:9",
    "resolution": "2k",
    "quality": "medium"
  }'

The response is the same task_id envelope as the standard channel — poll GET /api/v1/tasks/{task_id} and read result.images[].url.

Image-to-image

Supplying image_urls switches the request to editing. Describe the change in prompt; with several references, refer to them in order as <IMAGE_0>, <IMAGE_1>, and so on.

curl https://reapi.ai/api/v1/images/generations \
  -H "Authorization: Bearer rk_live_xxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "grok-imagine-image-2.0-official",
    "prompt": "Place the product from <IMAGE_0> on the marble surface in <IMAGE_1>, matching the window light",
    "image_urls": [
      "https://example.com/product.png",
      "https://example.com/scene.jpg"
    ],
    "aspect_ratio": "1:1",
    "resolution": "2k"
  }'

Reference images must be public http(s) URLs — base64 and data: URIs are rejected platform-wide. Each reference image is billed on top of the generated image.

Response

Submission returns the standard async envelope:

{
  "code": 200,
  "data": [
    {
      "status": "submitted",
      "task_id": "task_01K8AYYM6R03TGZ3Q2P0TZVNPX"
    }
  ]
}

Polled completion (GET /api/v1/tasks/{task_id}):

{
  "code": 200,
  "data": {
    "status": "completed",
    "result": {
      "images": [
        { "url": "https://cdn.reapi.ai/..." },
        { "url": "https://cdn.reapi.ai/..." }
      ]
    }
  }
}

result.images carries one entry per delivered image, so a request with n: 4 returns four entries. Image URLs are valid for 24 hours from completion — re-host the asset to your own storage if you need durable access.

Pricing

Both channels bill per delivered image, multiplied by n, but they dimension the rate differently:

• Standard — flat. Neither size nor resolution changes it, so a 16:9 image costs the same as a square one.
• Official — resolution and quality each move the rate and stack, for four combinations across three price points (1k+medium and 2k+low settle at the same rate). aspect_ratio and n do not change the per-image rate. Reference images are billed on top.

A request with n: 12 costs twelve times a single image on either channel.

credits = ceil(per_image_usd × n × 1000)

1 credit = $0.001. The rounding happens once on the aggregate, not per image. See the model page for the live rate.

Failed generations are refunded automatically — the reserved credits return to your balance when a task ends in failure.

Errors

grok-imagine-image-2-0 shares the platform-wide error envelope. The validation surface is small:

• prompt missing, empty, or whitespace-only → 400 INVALID_REQUEST.
• prompt longer than 10000 characters → 400 INVALID_REQUEST.
• n outside 1–12 → 400 INVALID_REQUEST.
• size outside the seven ratios and five pixel aliases → 400 INVALID_REQUEST.
• resolution set to anything other than quality → 400 INVALID_REQUEST.
• Any reference media (image_urls, image, mask) supplied → 400 INVALID_REQUEST with a text-to-image-only message.

On the official channel:

• aspect_ratio outside its 14 values → 400 INVALID_REQUEST. The standard channel's pixel aliases (1024x1024 etc.) are not valid here.
• resolution set to anything other than 1k / 2k → 400 INVALID_REQUEST. The standard channel's quality value is rejected.
• quality set to anything other than low / medium → 400 INVALID_REQUEST.
• More than 10 entries in image_urls, or a non-http(s) URL → 400 INVALID_REQUEST.

Tips

• The prompt is rendered as written, so lens, lighting and material language survives to the render. A thin prompt stays thin — there is no rewriting layer filling in composition for you.
• Ask for the shape you will ship. Generating 16:9 directly beats generating a square and cropping it, because the model composes for the frame.
• Batch exploration in one call. Twelve variations of one brief cost the same as twelve separate single-image requests but take one round trip.

Related

• imagen-4-0 — flat-rate text-to-image, one image per request
• seedream-5-0-lite — text-to-image plus image editing
• grok-imagine-1-0-video — the video sibling in the same family