# Nano Banana image editing and multi-image: inlineData and /v1/images/edits

Source: https://ofox.io/docs/develop/models/nano-banana/image-edit

For model IDs and choosing an endpoint, see the [Nano Banana overview](https://ofox.io/docs/develop/models/nano-banana).

**Key point**:The native API puts the image and the text in the same `parts`; the OpenAI-compatible API uses `/v1/images/edits`.

**Python (native)**

```python filename="edit.py"
from google import genai
from google.genai import types

client = genai.Client(
    api_key="YOUR_OFOX_API_KEY",
    http_options={"api_version": "v1beta", "base_url": "https://api.ofox.io/gemini"},
)

with open("apple.png", "rb") as f:
    image = types.Part.from_bytes(data=f.read(), mime_type="image/png")

resp = client.models.generate_content(
    model="google/gemini-3.1-flash-image",
    contents=["Make the apple green, keep everything else the same", image],
    config=types.GenerateContentConfig(
        response_modalities=["IMAGE"],
        image_config=types.ImageConfig(aspect_ratio="1:1", image_size="1K"),
    ),
)

for part in resp.candidates[0].content.parts:
    if part.inline_data:
        with open("apple_green.png", "wb") as f:
            f.write(part.inline_data.data)
```

**cURL (native)**

```bash filename="Terminal"
# The image's base64 is long; putting it straight on the command line fails with
# "Argument list too long", so write it to a file first
base64 < apple.png | tr -d '\n' > apple.b64
jq -n --rawfile img apple.b64 '{
  contents: [{ parts: [
    { text: "Make the apple green, keep everything else the same" },
    { inlineData: { mimeType: "image/png", data: $img } }
  ] }],
  generationConfig: {
    responseModalities: ["IMAGE"],
    imageConfig: { aspectRatio: "1:1", imageSize: "1K" }
  }
}' > body.json

curl 'https://api.ofox.io/gemini/v1beta/models/google/gemini-3.1-flash-image:generateContent' \
  --max-time 600 \
  -H "x-goog-api-key: $OFOX_API_KEY" \
  -H 'Content-Type: application/json' \
  -d @body.json \
  | jq -r '.candidates[0].content.parts[] | select(.inlineData) | .inlineData.data' | base64 --decode > apple_green.png
```

**cURL (OpenAI-compatible)**

```bash filename="Terminal"
curl https://api.ofox.io/v1/images/edits \
  --max-time 600 \
  -H "Authorization: Bearer $OFOX_API_KEY" \
  -F model=google/gemini-3.1-flash-image \
  -F prompt="Make the apple green, keep everything else the same" \
  -F image=@apple.png \
  | jq -r '.data[0].b64_json' | base64 --decode > apple_green.png
```

- **Multi-image composition**: put several `inlineData` items in the same `parts` and describe the role of each image in the text, e.g. "Place the person from the first image into the scene from the second image".
- **Keeping the aspect ratio**: pass `aspectRatio` when editing too; otherwise the model decides the output ratio, which may not match the original.
- **Reference image billing**: reference images are billed as input tokens. We measured 1,120 tokens for one 1024×1024 reference image; see the `IMAGE` entry of `promptTokensDetails` in the response.

If you run into errors, see [Troubleshooting errors](https://ofox.io/docs/develop/models/nano-banana/errors).