Grok Imagine Image Quality Text-to-Image API by xAI
xAI Grok Imagine generates polished visuals from natural-language prompts at 1K or 2K resolution, with 14 aspect ratios.
コード例
import requests
import time
# Step 1: Start image generation
generate_url = "https://api.atlascloud.ai/api/v1/model/generateImage"
headers = {
"Content-Type": "application/json",
"Authorization": "Bearer $ATLASCLOUD_API_KEY"
}
data = {
"model": "xai/grok-imagine-image-quality/text-to-image",
"prompt": "A beautiful landscape with mountains and lake",
"width": 512,
"height": 512,
"steps": 20,
"guidance_scale": 7.5,
}
generate_response = requests.post(generate_url, headers=headers, json=data)
generate_result = generate_response.json()
prediction_id = generate_result["data"]["id"]
# Step 2: Poll for result
poll_url = f"https://api.atlascloud.ai/api/v1/model/prediction/{prediction_id}"
def check_status():
while True:
response = requests.get(poll_url, headers={"Authorization": "Bearer $ATLASCLOUD_API_KEY"})
result = response.json()
if result["data"]["status"] == "completed":
print("Generated image:", result["data"]["outputs"][0])
return result["data"]["outputs"][0]
elif result["data"]["status"] == "failed":
raise Exception(result["data"]["error"] or "Generation failed")
else:
# Still processing, wait 2 seconds
time.sleep(2)
image_url = check_status()インストール
お使いの言語に必要なパッケージをインストールしてください。
pip install requests認証
すべての API リクエストには API キーによる認証が必要です。API キーは Atlas Cloud ダッシュボードから取得できます。
export ATLASCLOUD_API_KEY="your-api-key-here"HTTP ヘッダー
import os
API_KEY = os.environ.get("ATLASCLOUD_API_KEY")
headers = {
"Content-Type": "application/json",
"Authorization": f"Bearer {API_KEY}"
}API キーをクライアントサイドのコードや公開リポジトリに公開しないでください。代わりに環境変数またはバックエンドプロキシを使用してください。
リクエストを送信
import requests
url = "https://api.atlascloud.ai/api/v1/model/generateImage"
headers = {
"Content-Type": "application/json",
"Authorization": "Bearer $ATLASCLOUD_API_KEY"
}
data = {
"model": "your-model",
"prompt": "A beautiful landscape"
}
response = requests.post(url, headers=headers, json=data)
print(response.json())リクエストを送信
非同期生成リクエストを送信します。API は予測 ID を返し、それを使用してステータスの確認や結果の取得ができます。
/api/v1/model/generateImageリクエストボディ
import requests
url = "https://api.atlascloud.ai/api/v1/model/generateImage"
headers = {
"Content-Type": "application/json",
"Authorization": "Bearer $ATLASCLOUD_API_KEY"
}
data = {
"model": "xai/grok-imagine-image-quality/text-to-image",
"input": {
"prompt": "A beautiful landscape with mountains and lake"
}
}
response = requests.post(url, headers=headers, json=data)
result = response.json()
print(f"Prediction ID: {result['id']}")
print(f"Status: {result['status']}")レスポンス
{
"id": "pred_abc123",
"status": "processing",
"model": "model-name",
"created_at": "2025-01-01T00:00:00Z"
}ステータスを確認
予測エンドポイントをポーリングして、リクエストの現在のステータスを確認します。
/api/v1/model/prediction/{prediction_id}ポーリング例
import requests
import time
prediction_id = "pred_abc123"
url = f"https://api.atlascloud.ai/api/v1/model/prediction/{prediction_id}"
headers = { "Authorization": "Bearer $ATLASCLOUD_API_KEY" }
while True:
response = requests.get(url, headers=headers)
result = response.json()
status = result["data"]["status"]
print(f"Status: {status}")
if status in ["completed", "succeeded"]:
output_url = result["data"]["outputs"][0]
print(f"Output URL: {output_url}")
break
elif status == "failed":
print(f"Error: {result['data'].get('error', 'Unknown')}")
break
time.sleep(3)ステータス値
processingリクエストはまだ処理中です。completed生成が完了しました。出力が利用可能です。succeeded生成が成功しました。出力が利用可能です。failed生成に失敗しました。エラーフィールドを確認してください。完了レスポンス
{
"data": {
"id": "pred_abc123",
"status": "completed",
"outputs": [
"https://storage.atlascloud.ai/outputs/result.png"
],
"metrics": {
"predict_time": 8.3
},
"created_at": "2025-01-01T00:00:00Z",
"completed_at": "2025-01-01T00:00:10Z"
}
}ファイルをアップロード
Atlas Cloud ストレージにファイルをアップロードし、API リクエストで使用できる URL を取得します。multipart/form-data を使用してアップロードします。
/api/v1/model/uploadMediaアップロード例
import requests
url = "https://api.atlascloud.ai/api/v1/model/uploadMedia"
headers = { "Authorization": "Bearer $ATLASCLOUD_API_KEY" }
with open("image.png", "rb") as f:
files = {"file": ("image.png", f, "image/png")}
response = requests.post(url, headers=headers, files=files)
result = response.json()
download_url = result["data"]["download_url"]
print(f"File URL: {download_url}")レスポンス
{
"data": {
"download_url": "https://storage.atlascloud.ai/uploads/abc123/image.png",
"file_name": "image.png",
"content_type": "image/png",
"size": 1024000
}
}入力 Schema
以下のパラメータがリクエストボディで使用できます。
利用可能なパラメータはありません。
リクエストボディの例
{
"model": "xai/grok-imagine-image-quality/text-to-image"
}出力 Schema
API は生成された出力 URL を含む予測レスポンスを返します。
レスポンス例
{
"id": "pred_abc123",
"status": "completed",
"model": "model-name",
"outputs": [
"https://storage.atlascloud.ai/outputs/result.png"
],
"metrics": {
"predict_time": 8.3
},
"created_at": "2025-01-01T00:00:00Z",
"completed_at": "2025-01-01T00:00:10Z"
}Atlas Cloud Skills
Atlas Cloud Skills は 300 以上の AI モデルを AI コーディングアシスタントに直接統合します。ワンコマンドでインストールし、自然言語で画像・動画生成や LLM との対話が可能です。
対応クライアント
インストール
npx skills add AtlasCloudAI/atlas-cloud-skillsAPI キーの設定
Atlas Cloud ダッシュボードから API キーを取得し、環境変数として設定してください。
export ATLASCLOUD_API_KEY="your-api-key-here"機能
インストール後、AI アシスタントで自然言語を使用してすべての Atlas Cloud モデルにアクセスできます。
MCP Server
Atlas Cloud MCP Server は Model Context Protocol を通じて IDE と 300 以上の AI モデルを接続します。MCP 対応のあらゆるクライアントで動作します。
対応クライアント
インストール
npx -y atlascloud-mcp設定
以下の設定を IDE の MCP 設定ファイルに追加してください。
{
"mcpServers": {
"atlascloud": {
"command": "npx",
"args": [
"-y",
"atlascloud-mcp"
],
"env": {
"ATLASCLOUD_API_KEY": "your-api-key-here"
}
}
}
}利用可能なツール
APIスキーマ
スキーマが利用できません利用可能な例がありません
1. Introduction
Grok Imagine Image Quality is xAI's flagship image generation and editing system, also known as "Quality Mode," designed to deliver photorealistic imagery, legible in-image typography, and tight prompt adherence across diverse visual styles. This README applies to the following API model identifiers:
xai/grok-imagine-image-quality/text-to-imagexai/grok-imagine-image-quality/edit
Developed by xAI and built on the Aurora foundation—an autoregressive Mixture-of-Experts (MoE) architecture that differentiates it from diffusion-based competitors—Grok Imagine Image Quality targets creators, developers, and enterprises who require high-fidelity static imagery alongside natural-language editing. The consumer version launched on April 3, 2026 via grok.com/imagine and the Grok iOS/Android apps, and the API became publicly available on May 6, 2026 through the official announcement.
The system is exposed through two API variants that share the same underlying model but are optimized for distinct workflows. The xai/grok-imagine-image-quality/text-to-image endpoint produces images from text prompts with approximately 4-second latency, while xai/grok-imagine-image-quality/edit applies prompt-driven modifications to existing images—including multi-image reference composition—with approximately 13-second latency.
2. Key Features & Innovations
-
Aurora MoE Architecture: Unlike most image generators that rely on diffusion, Grok Imagine Image Quality is powered by Aurora, an autoregressive Mixture-of-Experts model. This approach yields strong facial consistency, accurate textures, and cinematic lighting behavior that reviewers have compared favorably with diffusion competitors on photorealistic sharpness.
-
High-Fidelity Text Rendering: The model produces legible in-image typography across multiple languages, addressing one of the historically weakest areas of generative image models. While Ideogram and GPT Image 2 still hold the lead in pure text rendering, Quality Mode closes the gap considerably versus prior Grok generations.
-
Prompt-Driven Editing Without Masks: The
xai/grok-imagine-image-quality/editvariant supports object addition, removal, swapping, style transfer, and multi-image reference composition entirely through natural-language prompts. No mask-based inpainting is required, and multi-turn iterative refinement is supported for progressive edits. -
Multi-Resolution and Multi-Format Output: Outputs are available at 1K (1024×1024) or 2K (2048×2048) resolution, across 13 aspect ratios ranging from 2:1 to 1:2. JPEG, PNG, and WebP formats are supported, with alpha channel available on PNG and WebP.
-
Batch Generation: Both variants accept a
num_imagesparameter (1–4) to generate multiple candidates per request, useful for creative exploration and A/B selection in production pipelines. -
Broad Stylistic Range: The model demonstrates competent prompt adherence across photorealistic, anime, oil painting, 3D-rendered, and abstract styles, making it suitable for varied creative and commercial briefs from a single endpoint.
-
Integrated Image-to-Video Pipeline: Grok Imagine Image Quality feeds directly into xAI's image-to-video capabilities, which currently rank #1 on the Artificial Analysis Image-to-Video Arena (Elo 1,336) and Multi-Image-to-Video Arena (Elo 1,342).
3. Model Architecture & Technical Details
Grok Imagine Image Quality uses the Aurora architecture—an autoregressive Mixture-of-Experts design. Rather than iteratively denoising latent representations as diffusion models do, autoregressive image models generate tokens sequentially, which contributes to the system's strong consistency across faces, fine textures, and typography. The MoE routing allows expert specialization across visual domains (portraiture, text, lighting, stylization) while keeping inference latency competitive.
Both API identifiers (xai/grok-imagine-image-quality/text-to-image and xai/grok-imagine-image-quality/edit) are served by the same underlying weights; the distinction lies in the input schema and conditioning path. The editing variant accepts a prompt plus one or more image_urls, enabling single-image edits as well as multi-image composition in which reference imagery informs the generated output.
API specifications:
| Parameter | Text-to-Image | Edit |
|---|---|---|
| Required inputs | prompt | prompt, image_urls |
num_images | 1–4 | 1–4 |
aspect_ratio | 13 options (2:1 to 1:2) | Defaults to auto |
resolution | 1k / 2k | 1k / 2k |
| Typical latency | ~4 s | ~13 s |
The model is positioned within xAI's tiered product line—Speed → Quality → Pro—where Quality Mode represents the balanced tier and Pro Mode adds 2K output with iterative editing workflows.
4. Performance Highlights
On the Artificial Analysis Text-to-Image Arena, Grok Imagine Image Quality sits within the top five models but trails the current leaders. Its strongest competitive results come from the image-to-video pipeline it feeds, where xAI's system ranks first overall.
Text-to-Image Arena (indicative rankings):
| Rank | Model | Developer | Elo Score |
|---|---|---|---|
| 1 | GPT Image 2 | OpenAI | 1338 |
| 2 | GPT Image 1.5 | OpenAI | 1273 |
| 3 | Nano Banana Pro | 1219 | |
| Top 5 | Grok Imagine Image Quality | xAI | Top-5 tier |
Image-to-Video / Multi-Image-to-Video Arena (pipeline context):
| Arena | Rank | Elo |
|---|---|---|
| Image-to-Video | #1 | 1,336 |
| Multi-Image-to-Video | #1 | 1,342 |
Qualitative strengths:
- Photoreal sharpness rated above Nano Banana by independent reviewers
- Strong facial consistency and cinematic lighting
- Competitive price-performance and fast inference
- Permissive content handling with an integrated video pipeline
Known limitations:
- In-image text rendering trails Ideogram, GPT Image 2, and FLUX
- Editing fidelity trails GPT Image 1.5 on complex structural edits
- Artistic stylization trails Midjourney V7 on illustrative aesthetics
- Moderation behavior has been reported as inconsistent by some users
5. Intended Use & Applications
-
Portrait and Character Art: The Aurora architecture's facial consistency and texture accuracy make
xai/grok-imagine-image-quality/text-to-imagewell suited for portrait generation, concept characters, and hero imagery where identity fidelity matters. -
Product and Commercial Marketing: Produce product advertisements, UGC-style marketing visuals, and product-film mockups at 2K resolution with cinematic lighting. The fast inference and per-image pricing support high-volume creative iteration.
-
Prompt-Driven Image Editing: Use
xai/grok-imagine-image-quality/editfor object addition, removal, swapping, and style transfer without requiring masks. Multi-turn refinement supports iterative polish workflows typical of design review cycles. -
Multi-Image Composition: The editing variant accepts multiple reference images, enabling workflows such as combining a subject with a new background, transferring wardrobe across references, or blending compositional cues from several inputs.
-
Social and Short-Form Content: Generate social-first imagery and stills that feed into the Grok Imagine image-to-video pipeline—currently ranked #1 on Artificial Analysis's video arenas—for an end-to-end static-to-motion workflow.
-
Concept Art and Creative Exploration: With batch sizes up to four images and broad stylistic range across photorealistic, anime, oil painting, 3D, and abstract styles, the model serves concept artists and creative directors exploring visual directions quickly.
-
Enterprise Creative Agencies and Media: The combination of 2K output, permissive content policy, and integrated video pipeline positions Grok Imagine Image Quality for creative agencies, entertainment and media production, and social-first consumer brands.
