> ## Documentation Index
> Fetch the complete documentation index at: https://docs.nexusapi.link/llms.txt
> Use this file to discover all available pages before exploring further.

# Image model guide

> How to call image models through NexusAPI.

Image models are called differently from text models. Before sending a request, use the [Model Marketplace](https://nexusapi.link/pricing) to confirm the model name, your API Key's group, and that the model is marked as supporting image generation.

<Warning>
  **Confirm the model capability first**

  Do not send image-generation parameters to a text chat endpoint, and do not use a text-model ID at an image endpoint. Retrying a mismatched model, endpoint, or parameter set will not solve the problem.
</Warning>

## Make a request

Image generation uses the OpenAI-style Images endpoint:

```text theme={"system"}
POST https://nexusapi.link/v1/images/generations
```

The minimum request needs only a model ID and prompt:

```bash theme={"system"}
curl 'https://nexusapi.link/v1/images/generations' \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer YOUR_API_KEY' \
  -d '{
    "model": "YOUR_IMAGE_MODEL_ID",
    "prompt": "A clean blue geometric technology company icon on a pure white background"
  }'
```

Replace `YOUR_IMAGE_MODEL_ID` with the exact image-capable model ID visible to your current group in the Model Marketplace.

## Parameters and results

Supported size, quality, image count, and response format vary by model. Send optional parameters such as `size`, `quality`, `n`, or `response_format` only when the Model Marketplace or [Apifox API reference](https://nexusapi.apifox.cn/) states they are supported. For shared authentication, model discovery, and protocol selection, see [API basics](/en/developer/api-basics).

Image results are usually returned in the `data` array. Whether a result is a URL or Base64 data, and how many images can be generated per request, depends on the selected model and its actual response.

## Waiting time and timeouts

The public `/v1/images/generations` endpoint is a **synchronous HTTP request path**: the client should wait for the result of this request. new-api rc4 does not provide a unified task-ID and polling API for generic image generation, and response formats can still differ by upstream provider.

* Image generation can take tens of seconds or longer. A short timeout in a client, corporate proxy, or browser extension can fail before the upstream responds.
* Do not submit the same image task repeatedly before confirming the outcome; this can cause duplicate images and duplicate charges.
* For `504`, `524`, or client timeouts, save the Request ID and time, then contact support for verification. Do not assume the task definitely succeeded or definitely did not run.

## Common issues

| Symptom                           | Check first                                                                                        |
| --------------------------------- | -------------------------------------------------------------------------------------------------- |
| `Image generation is not enabled` | Whether the model supports images, the image endpoint is used, and the current group can access it |
| `model not supported`             | The exact model ID and whether a text model was used at an image endpoint                          |
| `400` / invalid parameter         | Remove unverified optional parameters and try the minimum request first                            |
| `504` / `524` / client timeout    | Keep the Request ID and avoid an immediate duplicate submission                                    |
| No image result                   | Whether the client reads the `data` field and the returned format correctly                        |

When asking for help, provide the model name, Request ID, time, redacted request parameters, and full error text. Never provide your complete API Key.

## Next steps

* [Models, groups, and pricing](/en/billing/models-pricing)
* [Choose an integration protocol](/en/guide/integration-overview)
* [Common API errors](/en/faq/common-errors)