Gemini 2.5 Flash Image parameters

The prompt and optional source image belong in contents[].parts[]. This model uses the shared Gemini-native body, with one important YingTu rule: omit generationConfig.imageConfig.imageSize.

Field	Value for this model
Model path	`gemini-2.5-flash-image`
`responseModalities`	Include `IMAGE`; `TEXT` is optional
`aspectRatio`	Optional; `1:1` was verified live
`imageSize`	Omit
`candidateCount`	Omit or set to `1`
Top-level `n`	Unsupported
YingTu price	USD 0.020 per request

{
  "contents": [{
    "role": "user",
    "parts": [{"text": "A compact travel poster with one clear focal point"}]
  }],
  "generationConfig": {
    "responseModalities": ["TEXT", "IMAGE"],
    "imageConfig": {"aspectRatio": "1:1"}
  }
}

Editing input

Add the source image after the instruction in the same parts array. Provide raw Base64 bytes without a data-URL prefix and use the source file’s real MIME type.

{
  "contents": [{
    "role": "user",
    "parts": [
      {"text": "Keep the subject and replace the background with warm paper"},
      {"inlineData": {"mimeType": "image/png", "data": "BASE64_SOURCE_IMAGE"}}
    ]
  }],
  "generationConfig": {
    "responseModalities": ["TEXT", "IMAGE"],
    "imageConfig": {"aspectRatio": "1:1"}
  }
}

Do not add imageSize: "1K" merely because the output is approximately 1K. Current YingTu parameter rules require that field to be absent for this model.

The verified responses used role: "model", finishReason: "STOP", and PNG image data. Encoding is not guaranteed; always read inlineData.mimeType.

​Editing input

Editing input