Generate
Edit
Customize
“Cinematic photo of a person in a linen jacket”
Generate
Edit
Customize
AI Image & Video Generation API
40+ image and video models, one REST API. Generate, upscale, train custom LoRAs, and build creative pipelines. Simple compute-unit pricing. No infrastructure to manage.
# Step 1: Create generation job
curl -X POST https://api.krea.ai/generate/image/bfl/flux-1-dev \
-H "Authorization: Bearer YOUR_API_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"prompt": "a serene mountain landscape at sunset",
"width": 1024,
"height": 576,
"steps": 28
}'
# Response: {"job_id": "550e8400-e29b-41d4-a716-446655440000", ...}
# Step 2: Poll for completion (repeat until completed)
curl -X GET https://api.krea.ai/jobs/550e8400-e29b-41d4-a716-446655440000 \
-H "Authorization: Bearer YOUR_API_TOKEN"Simple integration
Start generating in minutes
One POST request. One prompt. Get an image back. Works in any language you already use.
# Step 1: Create generation job
curl -X POST https://api.krea.ai/generate/image/bfl/flux-1-dev \
-H "Authorization: Bearer YOUR_API_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"prompt": "a serene mountain landscape at sunset",
"width": 1024,
"height": 576,
"steps": 28
}'
# Response: {"job_id": "550e8400-e29b-41d4-a716-446655440000", ...}
# Step 2: Poll for completion (repeat until completed)
curl -X GET https://api.krea.ai/jobs/550e8400-e29b-41d4-a716-446655440000 \
-H "Authorization: Bearer YOUR_API_TOKEN"{
"job_id": "550e8400-e29b-41d4-a716-446655440000",
"status": "completed",
"created_at": "2025-01-15T10:30:00.000Z",
"completed_at": "2025-01-15T10:30:04.000Z",
"result": {
"urls": [
"https://krea.ai/generations/your-image.png"
]
}
}
Text to Image API
Generate product shots, marketing assets, and editorial photography from a single API call. Pick from 20+ models like Flux, Imagen 4, and Nano Banana Pro, or let Krea choose the best one for your prompt. No need to manage multiple provider accounts or stitch together different SDKs.
Get Started
Video Generation API
Turn text or images into video with models like Veo 3, Kling, and Wan 2.5. Build video ads, social content, or product demos programmatically. Many of these models have limited public API access elsewhere. On Krea, they all share the same endpoint.
Get Started
Image Upscaling API
Take any image up to 22K resolution with Topaz-powered AI enhancement. Three modes: standard for clean upscaling, bloom for stylized detail, and generative for adding new detail at scale. Perfect for print-ready assets or high-res product imagery.
Get Started
LoRA Training API
Train custom models on your brand, characters, or product line. Upload images, fine-tune on Flux, and use your trained styles in every generation request. Build pipelines that produce on-brand assets without manual intervention.
Get StartedBuild workflows visually, ship them as an API
Chain models together in a no-code node editor. Connect image generation to upscaling, feed outputs into video, or run batch jobs with branching logic. When it works, hit deploy and call it from your code like any other endpoint.
Get StartedThe best AI models, already picked for you
40+ models from Flux, Imagen, Veo, Kling, and more. We test, curate, and maintain them so you skip the guesswork. One integration, one billing system. Swap models per request.
Image Models
Nano Banana 2
$0.06 per image
Nano Banana Pro
$0.15 per image
Flux
$0.04 per image
Flux 1.1 Pro
$0.06 per image
Imagen 4
$0.04 per image
Imagen 4 Ultra
$0.06 per image
ChatGPT Image
from $0.03 per image
Seedream 5 Lite
$0.04 per image
Ideogram 3.0
$0.06 per image
Flux Kontext
$0.04 per image
Video Models
Veo 3
$0.20 per second of video
Veo 3.1
$0.20 per second of video
Kling 2.6
$0.07 per second of video
Hailuo 2.3
from $0.28 per video
Runway Gen-4.5
$0.12 per second of video
Sora 2
$0.10 per second of video
Wan 2.5
from $0.05 per second of video
Seedance Pro
from $0.25 per video
Ray 2
$0.40 per video
Veo 2
$0.50 per second of video
Enhancement Models
Topaz
$0.10 per upscaled image
Topaz Generative
$0.27 per upscaled image
Topaz Bloom
$0.51 per upscaled image
Developer experience
Built to get out of your way
Multi-language SDKs
Full examples in Python, Node.js, Go, and cURL. OpenAPI spec included for code generation.
Webhooks
Add an X-Webhook-URL header and get a POST when your job finishes. No polling required.
Automatic queueing
Hit your concurrency limit? Jobs queue in a backlog automatically instead of getting dropped.
Bearer token auth
Create API keys in your workspace settings. Pass them as Bearer headers. That's it.
Asset management
Upload, list, and retrieve generated assets through the API. Up to 75 MB per file.
Job lifecycle tracking
Every job goes through queued, processing, and completed states. Poll or use webhooks to track.
One price covers the app and the API. No extra subscription. Pay only for what you generate.
Free
Get free daily credits to try basic features.
Basic
Access our most popular features
Pro
Advanced features and discounts on compute units
Max
Full access with higher discounts on compute units
For Teams and Enterprises
Workplace management, collaboration, and enterprise customizations
Business
Secure and collaborative workspace for growing teams
Enterprise
Enterprise-grade security with dedicated support and admin features
Frequently Asked Questions
It depends on what you're building. Some models are faster, some are more photorealistic, and some handle text in images better. Krea gives you access to 20+ image models through a single endpoint, so you can test and swap models per request without changing your integration. Most teams start with Flux for speed and switch to Imagen 4 or Nano Banana Pro when they need higher fidelity.
Yes. Krea includes compute units on the free tier, so you can make API calls without a credit card. The free plan gives you enough to prototype your integration and test different models. When you're ready for production volume, upgrade to a paid plan for more compute and higher concurrency.
Everything runs on compute units. Each model has a fixed cost per generation (Flux is ~5 CU, Veo 3 is ~1,017 CU). Your API calls and the Krea web app share the same pool. No separate API subscription, no per-model fees, and failed or cancelled jobs never cost you anything.
Sign up at krea.ai, go to workspace settings and create an API key. Make a POST to the generation endpoint with your prompt and model. You'll get a job ID back. Poll for the result, or set a webhook URL to get notified when it's done. The quickstart guide walks you through it in about 5 minutes.
Replicate and Fal are inference platforms. You bring your own model or pick from a catalog and pay per GPU-second. Krea is different in a few ways: we curate 40+ production-ready models so you don't have to evaluate quality yourself, billing is per-generation compute units (not GPU time), failed jobs are never charged, and you get a full creative web app on top of the API. If your team uses the Krea app for visual work, the API shares the same account and credits.
Yes. Midjourney doesn't offer a public API, which makes it impossible to integrate into apps or automate workflows. Krea gives you REST API access to 20+ image models with quality on par or better, plus LoRA training, upscaling, and video generation. If you've been looking for a way to build on top of Midjourney-quality output, this is it.
Yes. The API gives you access to 20+ video models, including Sora 2, Veo 3, Kling, Hailuo, and Runway Gen-4. Generate text-to-video and image-to-video clips from 2 to 12 seconds. Several of these models have limited public API availability elsewhere.
40+ across image generation (Flux, Imagen 4, ChatGPT Image, Nano Banana Pro, Seedream, Ideogram), video generation (Veo 3, Kling, Hailuo, Runway, Sora 2, Wan, Seedance), and image enhancement (Topaz, Topaz Bloom, Topaz Generative). New models get added regularly. Check the full model list for current availability.
You submit a request and get a job ID back immediately. The job moves through queued, processing, and completed (or failed/cancelled). Poll the status endpoint every 2-5 seconds, or pass a webhook URL in the request header to get notified automatically. Failed and cancelled jobs don't consume compute units.
Yes. Upload training images and create fine-tuned LoRAs for styles, characters, or subjects. Trained models work with Flux, Wan, and more. Use them in subsequent generation requests to keep output on-brand without manual work. See the training docs for details.
Start building with the Krea API
Get your API key and start generating with 40+ AI models in minutes. Same compute units, same account, one API.
Get Started