**Cost**: Model-priced per successful call. See the Credit Matrix for `image_remix` model and resolution prices.

Image Remix supports two input modes:

1. **JSON URL mode**: send `application/json` with the existing URL field.
2. **Direct upload mode**: send `multipart/form-data` with a local image file in the `image` field. The compatible field name `file` is also accepted for the primary image. Codia checks credits before reading and uploading the file, so insufficient-credit requests return `402` without uploading anything.

Direct upload example:

```bash
curl 'https://openapi.codia.ai/v2/open/image/remix' \
  -H 'Authorization: Bearer {codia_api_key}' \
  -F 'image=@./image.png' \
  -F 'prompt=turn into pixel art' \
  -F 'size=1024x1024'
```

Generate a new image that **preserves the structure** of a source (composition, layout, pose, key shapes) while changing the rest according to a prompt. Stronger structural preservation than `image_to_image` — use this when the silhouette / framing matters and only the surface should change.

## What this API is for

Thumbnail variants that keep the same crop; alternate color schemes for the same composition; style variants for marketing across regions; iterating on a moodboard image.

## Request

| Field | Required | Notes |
|---|---|---|
| `image_url` | yes | Public URL of the source. |
| `prompt` | yes | Description of the desired remix. |
| `model` | no | Default `nano_banana_2`. Also: `nano_banana_pro`, `flux_3`, `ideogram_v4`. |
| `size` | no | Output dimensions as `WIDTHxHEIGHT`. If omitted, output matches the source's aspect. |

## Response

Shared `ImageUrlListSuccess` — one URL in `data.image_urls`.

## Billing & limits

- Credits follow the `image_remix` model/resolution table in the global Credit Matrix.
- Synchronous.

## Errors

Standard set.

> **Sync vs. async**
> This synchronous endpoint returns the result inline — the simplest path. For production pipelines that need webhook delivery, automatic retry with idempotency, or explicit task cancellation, call `POST /v2/open/tasks` with `operation=image_remix` instead. Inputs and credit cost are identical.

可选，由用户自定义的幂等键（最多 255 个字符）。设置后，成功（2xx）的响应会缓存 7 天。使用相同的键 + 相同的请求体重试，会返回缓存的响应（不额外扣除积分）；使用相同的键但不同的请求体重试，则返回 409。用于确保 POST 请求在网络故障时可以安全重试。对于 multipart/form-data 上传，请求会按表单字段和文件内容进行比较，因此即使客户端生成了新的 multipart boundary，重试的上传也会被重放。

Idempotency-Key

image_url

prompt

执行 Remix 的模型。默认为 `nano_banana_2`。积分按 `image_remix` 模型/分辨率表计算。

| 模型 | 默认 | 1K | 2K | 4K |
| --- | ---: | ---: | ---: | ---: |
| `nano_banana_2` | 18 | 18 | 27 | 41 |
| `nano_banana_pro` | 36 | 36 | 36 | 65 |
| `gpt_image` (medium) | 15 | 15 | 29 | 48 |
| `gpt_image_2_5` (medium) | 4 | 4 | 8 | 12 |
| `seedream_5_pro` | 12 | 12 | 24 | - |
| `seedream_5_flash` | 5 | 5 | 5 | - |
| `seedream_5` | 10 | 10 | 10 | 10 |
| `seedream_4_5` | 11 | 11 | 11 | 11 |
| `recraft_v4` | 11 | 11 | 11 | - |
| `flux_3` | 13 | 13 | 27 | 162 |
| `flux_2_pro` | 8 | 8 | 20 | - |
| `flux_2_max` | 19 | 19 | 43 | - |
| `ideogram_v4` | 16 | 16 | 16 | 16 |

GPT 行显示的是 medium 质量。其他质量档位请参阅上方共享的 GPT Image 积分矩阵。

model

输出图片尺寸，格式为 `WIDTHxHEIGHT`。FLUX 2 Pro/Max 也接受 1K/2K 尺寸预设。超大请求（包括 4K）会被缩减到最多约 4MP（方形图片为 2048x2048）；不支持原生 4K 输出。计费按调整后的输出面积计算：总像素不超过 1,048,576 的按 1K 价格计费，更大的输出按 2K 价格计费。显式指定的 size 优先于 quality 中的分辨率提示。

size

对于 `gpt_image`，可使用 `low`、`medium`、`high` 或 `auto`。对于 `gpt_image_2_5`（包括 Flare 和 Sunburst），还支持 `xhigh` 和 `max`。省略 quality 或设为 `auto` 时，按 medium 计费档位计费。使用 `size` 指定输出尺寸。

quality

image

兼容用的源图片文件字段（已弃用：请改用 `image`）。

file

code

message

image_urls

data

请求体格式错误或缺少必填参数

Idempotency-Key 已被用于不同的请求体。

请求频率超限。请等待 `Retry-After` 响应头中指定的秒数后再重试。

保留原图的构图和形状，根据提示词重新风格化，生成一张新图片。

图片重混

输入

输出

相关 API