Why Nebius AI Studio stands outWhy Nebius AI Studio stands out

We’ve built our service to scale with your imagination, with no resolution caps or strict quotas.

🔍 Superior Resolution & Quality

• Generate images up to 2000×2000 pixels (4 megapixels) — that’s up to 4 times larger than most alternatives

• Adjust your results with advanced parameter controls

• Achieve consistent results with seed management and negative prompts

💰 Industry-leading pricing

• No hidden fees or size-based pricing variations

• Simple, predictable costs for easier budgeting

🚀 Built for production

• OpenAI-compatible API for seamless integration

• Flexible rate limits that grow with your needs

• Enterprise-grade infrastructure for reliable scaling

Short, hands-on examplesShort, hands-on examples

Implementing our text-to-image service is straightforward. Here’s a quick example using cURL:

curl -X 'POST' \
  'https://api.studio.nebius.ai/v1/images/generations' \
  -H 'accept: application/json' \
  -H "Authorization: Bearer $NEBIUS_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "black-forest-labs/flux-schnell",
    "prompt": "A serene landscape at sunset, with rolling hills and a lone tree silhouette",
    "response_format": "url"
  }'

And here’s the one for Python:

from openai import OpenAI

client = OpenAI(
  base_url="https://api.studio.nebius.ai/v1/",
  api_key="your_api_key"
)

response = client.images.generate(
  model="stability-ai/sdxl",
  prompt="A serene mountain landscape at sunrise, with mist in the valley",
  size="1024x1024",
  response_format="b64_json",
  extra_body={
    "response_extension": "webp",  # Optimize for web delivery
    "num_inference_steps": 40,     # Higher quality output
    "seed": 123456,                # Reproducible results
    "negative_prompt": "dark, cloudy, storm, hazy"
  }
)

Key parameters to customize your generations include:

• size: Support for resolutions up to 2048×2048 pixels

• num_inference_steps: Adjust the generation quality

• seed: Enable consistent, reproducible results

• negative_prompt: Refine outputs by specifying what to exclude

• response_format: Get results as URLs or base64-encoded images

All models available in Studio support these parameters, allowing you to optimize for your specific needs — whether it’s ultra-fast generation with Flux Schnell or premium quality with SDXL.

Want to experiment without coding? Try our interactive Playground to test different models and prompts instantly.

Three ways to master prompt craftingThree ways to master prompt crafting

The key to exceptional AI-generated images is in how you describe them. Here are three proven techniques to improve your results:

1. Be specific with composition. Instead of “A cat,” try “A Persian cat sitting in a sunlit window, soft morning light, professional photography.”

2. Use negative prompts to refine results:

{
"prompt": "Professional headshot of a business person",
"negative_prompt": "blurry, distorted features, unrealistic lighting"
}

3. Save seeds for consistency. Keep your seed value when you find the perfect style and use it across multiple generations for consistent results. This approach is perfect for creating sets of related images.

ConclusionConclusion

What’s nextWhat’s next

We’re actively developing image-to-image transformations, advanced editing capabilities, model fine-tuning options, enhanced control features and an ability of running LoRA models. Plus, we’re continuously optimizing the models for even better quality and performance.

How to get started in minutesHow to get started in minutes

• Experiment with our Playground for instant results.

• Get your API key to integrate text-to-image into your workflows.

• Explore our documentation for deeper insights.

We can’t wait to see what you create.