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

# make_image

> Generate or edit images from a prompt and optional reference images.

```
POST /api/v1/tools/make_image/run
```

Generates images from a text prompt. With `reference_images` attached it edits/composes from them (put a product on a model, restyle a photo, keep a face consistent); without references it generates purely from text.

## Routing

The default `model: "auto"` routes by whether references are attached:

* **With `reference_images`** → `nano-banana-2` (reference-driven editing/composition).
* **Without** → `gpt-image-2` (text-to-image).

You can pin `model` explicitly, with one hard rule: **`nano-banana-2` requires at least one reference image** — it is an edit model and a ref-less request to it is rejected with a `400`.

## Arguments

| Argument           | Type                    | Required  | Default  | Description                                                                                                                                                                                                                                                        |
| ------------------ | ----------------------- | --------- | -------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| `prompt`           | string                  | yes       | —        | What to generate, or the edit to apply to the references.                                                                                                                                                                                                          |
| `reference_images` | array of strings (URLs) | no        | —        | Up to **14** publicly fetchable image URLs. Attaching any requires `owns_references: true`.                                                                                                                                                                        |
| `owns_references`  | boolean                 | with refs | —        | Likeness-consent attestation, identical to [`make_video`](/tools/make-video#owns_references-consent): required `true` when references are attached, otherwise `400` with no debit; persisted on the run with a timestamp as an audit trail.                        |
| `model`            | string                  | no        | `"auto"` | One of `auto`, `nano-banana-2`, `gpt-image-2`.                                                                                                                                                                                                                     |
| `aspect_ratio`     | string                  | no        | `"9:16"` | One of `1:1`, `16:9`, `9:16`, `4:3`, `3:4`, `3:2`, `2:3`, `21:9`.                                                                                                                                                                                                  |
| `resolution`       | string                  | no        | `"1k"`   | One of `1k`, `2k`, `4k`. Higher resolutions cost more — see [pricing](/credits-and-refunds#what-runs-cost).                                                                                                                                                        |
| `count`            | number                  | no        | `1`      | Number of images to generate, 1–4 — all are variations of the *same* prompt. For distinct scenes or outfits, submit one run per prompt. Cost is per image.                                                                                                         |
| `enhance_prompt`   | boolean                 | no        | `false`  | Rewrite a rough prompt via a prompt-writing model before generation (+3 credits); both prompts are stored and searchable. Same semantics as [`make_video`](/tools/make-video#enhance_prompt).                                                                      |
| `influencer`       | string                  | no        | —        | Slug of a saved [influencer](/tools/create-influencer). The server attaches its portrait + character sheet as references and prepends its appearance description to your prompt, keeping the same character across generations. An unknown slug → `400`, no debit. |

## Example — text-to-image

```bash theme={null}
curl -X POST https://getagenthook.com/api/v1/tools/make_image/run \
  -H "Authorization: Bearer $AGENTHOOK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Product shot of a matte-black water bottle on a concrete plinth, soft studio light, pale lime accent backdrop",
    "aspect_ratio": "1:1",
    "count": 2
  }'
```

## Example — reference-driven edit

```bash theme={null}
curl -X POST https://getagenthook.com/api/v1/tools/make_image/run \
  -H "Authorization: Bearer $AGENTHOOK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Put the person from image 1 in a denim jacket, golden-hour street background",
    "reference_images": ["https://example.com/my-face.jpg"],
    "owns_references": true,
    "resolution": "2k"
  }'
```

## Response

`202` with `run_id` and `credits_charged`; poll `GET /api/v1/runs/{run_id}`. On `completed`, `output` is an array of durable image URLs — one per `count`. On `failed`, credits are [refunded instantly](/credits-and-refunds#instant-refunds-on-failure).