An image generation API creates images from a text prompt, and on many models edits or combines images you supply. You send the prompt with settings such as aspect ratio and resolution, and the API returns the finished images as URLs or base64 data, billed per image.
An image generation API turns a written description into a picture. You send a prompt, for example a ceramic mug on a marble counter in soft morning light, and the model returns one or more new images. This is called text-to-image.
Many image models also take images as input. With editing, you send a source image and an instruction, such as a new background or a different color, and the model returns the edited image. With reference images, the model combines subjects, products or styles from several pictures into a new one. Depending on the model, you can also request a transparent background, a set of related images in one request, or a split of one image into separate editable layers.
You send POST /v1/images/generations with the model ID, the prompt and optional settings such as aspect_ratio, resolution and num_images. To edit, add the source or reference images as URLs in image. The API waits for the generation and returns the finished images as URLs, or as base64 data when you set response_format to b64_json.
If a generation is still running after a long wait, the response carries a job ID instead, and you poll GET /v1/jobs/{job_id} for the images. To receive a job ID at once, for example when you queue many generations, send the header Prefer: respond-async.
Side-by-side model comparisons show two models' capabilities and rates together.
Browse the image models and try your prompts in the Playground. When you are ready to build, create an API key in the dashboard and call https://api.empiriolabs.ai/v1/images/generations. Each model page lists the parameters the model accepts, and the image generation reference covers the request and the response in full.
Pay as you go, or choose a plan with weekly allowances. Current per-image rates are on the pricing page.
18 models, cheapest first where they share a billing unit. Each model's page has its full rate card, parameters and code.
Endpoints
/v1/images/generations POST /v1/images/editsPriced per image
| Model | Lowest rate | Applies to |
|---|---|---|
| FLUX.2 Klein 4B Black Forest Labs | $0.014$0.0085 per image | Image generation |
| GPT Image 2 OpenAI | $0.012 per image | Low |
| MOSS-Image Turbo OpenMOSS | $0.05$0.015 per image | Image generation |
| Qwen Image 2.1 Alibaba Cloud | $0.016 per image | Turbo |
| Wan2.1 Image Alibaba Cloud | $0.025 per image | Turbo |
| Wan2.2 Image Alibaba Cloud | $0.025 per image | Flash |
| Qwen Image 3.0 Alibaba Cloud | $0.030 per image | Base Output (1K or 2K) |
| Seedream 5.0 Flash ByteDance | $0.030 per image | Image generation |
| Model | Lowest rate | Applies to |
|---|---|---|
| Wan2.5 Image Alibaba Cloud | $0.030 per image | Standard |
| Wan2.7 Image Alibaba Cloud | $0.030 per image | Standard |
| Qwen Image 2.0 Alibaba Cloud | $0.035$0.0322 per image | Standard |
| Seedream 5.5 ByteDance | $0.0332 per image | Output up to 2.61MP |
| Seedream 5.0 Lite ByteDance | $0.0350 per image | Standard |
| Grok Imagine Image 2.0 xAI | $0.048 per image | Low quality, 1K |
| Seedream 4.0 ByteDance | $0.060 per image | Standard |
| Seedream 5.0 Pro ByteDance | $0.075 per image | Output up to 2.61MP |
| Seedream 4.5 ByteDance | $0.080 per image | Standard |
| Hunyuan Image 3 Tencent | $0.13 per image | Standard |
Text-to-image creates a new picture from a prompt alone. Image editing starts from one or more images you supply and changes them according to your instruction, such as a new background or an added object.
Image models are billed per generated image, and the rate often depends on the quality or resolution tier. Some models also bill each input image. On EmpirioLabs AI you can pay as you go or choose a plan with weekly allowances.
Yes, on models that support it. They return a PNG with an alpha channel, which you can place over any background.
Use a model that accepts a seed, and send the same seed, prompt and settings again. Without a fixed seed, each request produces a different image.
FLUX.2 Klein 4B lists the lowest rate per image of the 18 Image Generation models priced that way, followed by GPT Image 2, MOSS-Image Turbo. The price table on this page shows each model's lowest rate and what it applies to.
Check out our pricing or reach out if you want your own model deployed on our stack.