从 Imagen 迁移到 Gemini Nano Banana 模型


所有 Imagen 模型均已弃用，最早将于 2026 年 8 月 17 日关停。此弃用和关停适用于整个 Google 以及 Gemini Developer API 和 Agent Platform Gemini API (formerly Vertex AI)。

在此关停日期之前，为避免服务中断，您应按照本指南中的说明，将应用从使用 Imagen 模型迁移到使用 Gemini Nano Banana 模型。

如果您遇到与此弃用和关停相关的紧急问题，请与 Firebase 支持团队联系。

Nano Banana 替换模型

下表可帮助您为应用选择替代 Nano Banana 模型。

Imagen 个模型 Gemini Nano Banana 个模型
imagen-4.0-fast-generate-001 gemini-nano-banana-2.1（思考等级为MINIMAL）
imagen-4.0-generate-001 gemini-nano-banana-2.1（思考等级为HIGH）
imagen-4.0-ultra-generate-001 gemini-3-pro-image
imagen-3.0-capability-001 gemini-nano-banana-2.1

迁移您的应用

本部分展示了从 Imagen 模型迁移到 Gemini Nano Banana 模型的示例。

根据文字生成图片

点击您的 Gemini API 提供商，以查看此页面上特定于提供商的内容和代码。

如需根据文本生成图片，请通过做出以下更改来迁移应用：

  • 使用适当的替代 Nano Banana 模型（例如 gemini-nano-banana-2.1）。

  • 创建 GenerativeModel 实例（而不是 ImagenModel 实例）。

  • 更新模型配置选项以适应 Nano Banana 模型。

    • 在此配置中，将回答模态设置为 IMAGE。
      请注意，Nano Banana 模型可以配置为同时返回图片和文本。

Swift

之前


import FirebaseAILogic

// Initialize the Gemini Developer API backend service.
let ai = FirebaseAI.firebaseAI(backend: .googleAI())

// Create an `ImagenModel` instance with a model that supports your use case.
let model = ai.imagenModel(modelName: "IMAGEN_MODEL_NAME")

// Provide an image generation prompt.
let prompt = "An astronaut riding a horse"

// To generate an image, call `generateImages` with the text prompt.
let response = try await model.generateImages(prompt: prompt)

// Handle the generated image.
guard let image = response.images.first else {
  fatalError("No image in the response.")
}
let uiImage = UIImage(data: image.data)

之后


import FirebaseAILogic

// Initialize the Gemini Developer API backend service.
let ai = FirebaseAI.firebaseAI(backend: .googleAI())

// Create a `GenerativeModel` instance with a Gemini model that supports image output.
let model = ai.generativeModel(
  modelName: "GEMINI_NANOBANANA_MODEL_NAME",
  generationConfig: GenerationConfig(
    responseModalities: [.image],
    imageConfig: ImageConfig(aspectRatio: .landscape4x3)
  )
)

// Provide an image generation prompt.
let prompt = "An astronaut riding a horse"

// To generate an image, call `generateContent` with the text prompt.
let response = try await model.generateContent(prompt)

// Handle the case where no images were generated.
guard let inlineDataPart = response.inlineDataParts.first else {
  fatalError("No image in the response.")
}

// Process the image.
guard let uiImage = UIImage(data: inlineDataPart.data) else {
  fatalError("Failed to convert data to UIImage.")
}

Kotlin

之前


// Initialize the Gemini Developer API backend service.
val ai = Firebase.ai(backend = GenerativeBackend.googleAI())

// Create an `ImagenModel` instance with an Imagen model that supports your use case.
val model = ai.imagenModel("IMAGEN_MODEL_NAME")

// Provide an image generation prompt.
val prompt = "An astronaut riding a horse"

// To generate an image, call `generateImages` with the text prompt.
val imageResponse = model.generateImages(prompt)

// Handle the generated image.
val image = imageResponse.images.first()

val bitmapImage = image.asBitmap()

之后


// Initialize the Gemini Developer API backend service.
val ai = Firebase.ai(backend = GenerativeBackend.googleAI())

// Create a `GenerativeModel` instance with a Gemini model that supports image output.
val model = ai.generativeModel(
    modelName = "GEMINI_NANOBANANA_MODEL_NAME",
    generationConfig = generationConfig {
      responseModalities = listOf(ResponseModality.IMAGE),
      imageConfig = imageConfig {
        aspectRatio = AspectRatio.LANDSCAPE_4x3
      }
    }
)

// Provide an image generation prompt.
val prompt = "An astronaut riding a horse"

// To generate an image, call `generateContent` with the text prompt.
val imageResponse = model.generateContent(prompt)

if (imageResponse.finishReason == FinishReason.NO_IMAGE) {
  // Handle the case where no images were generated.
} else {
  // Handle the generated image.
  val bitmapImage = imageResponse.candidates.first().content.parts.filterIsInstance().firstOrNull()?.image
}

Java

之前


// Initialize the Gemini Developer API backend service.
// Create an `ImagenModel` instance with an Imagen model that supports your use case.
ImagenModel imagenModel = FirebaseAI.getInstance(GenerativeBackend.googleAI())
        .imagenModel(
                /* modelName */ "IMAGEN_MODEL_NAME");

ImagenModelFutures model = ImagenModelFutures.from(imagenModel);

// Provide an image generation prompt.
String prompt = "An astronaut riding a horse";

// To generate an image, call `generateImages` with the text prompt.
Futures.addCallback(model.generateImages(prompt), new FutureCallback<ImagenGenerationResponse>() {
    @Override
    public void onSuccess(ImagenGenerationResponse result) {
        if (result.getImages().isEmpty()) {
            Log.d("TAG", "No images generated");
        }
        Bitmap bitmap = result.getImages().get(0).asBitmap();
        // Use the bitmap to display the image in your UI.
    }

    @Override
    public void onFailure(Throwable t) {
        // ...
    }
}, Executors.newSingleThreadExecutor());

之后


// Initialize the Gemini Developer API backend service.
// Create a `GenerativeModel` instance with a Gemini model that supports image output.
GenerativeModel ai = FirebaseAI.getInstance(GenerativeBackend.googleAI()).generativeModel(
    "GEMINI_NANOBANANA_MODEL_NAME",
    new GenerationConfig.Builder()
        .setResponseModalities(Arrays.asList(ResponseModality.IMAGE))
        .setImageConfig(new ImageConfig(AspectRatio.LANDSCAPE_4x3, null))
        .build()
);

GenerativeModelFutures model = GenerativeModelFutures.from(ai);

// Provide a text prompt instructing the model to generate an image.
Content prompt = new Content.Builder()
        .addText("An astronaut riding a horse")
        .build();

// To generate an image, call `generateContent` with the text input.
Executor executor = Executors.newSingleThreadExecutor();
ListenableFuture response = model.generateContent(prompt);
Futures.addCallback(response, new FutureCallback() {
    @Override
    public void onSuccess(GenerateContentResponse result) {
        if (result.finishReason == FinishReason.NO_IMAGE) {
            // handle the case where no images were generated
            return;
        }
        // iterate over all the parts in the first candidate in the result object.
        for (Part part : result.getCandidates().get(0).getContent().getParts()) {
            if (part instanceof ImagePart) {
                ImagePart imagePart = (ImagePart) part;
                // The returned image as a bitmap
                Bitmap generatedImageAsBitmap = imagePart.getImage();
                break;
            }
        }
    }

    @Override
    public void onFailure(Throwable t) {
        t.printStackTrace();
    }
}, executor);

Web

之前


import { initializeApp } from "firebase/app";
import { getAI, getGenerativeModel, getImagenModel, GoogleAIBackend } from "firebase/ai";

// TODO(developer): Replace the following with your app's Firebase configuration
const firebaseConfig = {
  // ...
};

// Initialize FirebaseApp
const firebaseApp = initializeApp(firebaseConfig);

// Initialize the Gemini Developer API backend service.
const ai = getAI(firebaseApp, { backend: new GoogleAIBackend() });

// Create an `ImagenModel` instance with an Imagen model that supports your use case.
const model = getImagenModel(ai, { model: "IMAGEN_MODEL_NAME" });

// Provide an image generation prompt.
const prompt = "An astronaut riding a horse.";

// To generate an image, call `generateImages` with the text prompt.
const response = await model.generateImages(prompt)

// If fewer images were generated than were requested,
// then `filteredReason` will describe the reason they were filtered out.
if (response.filteredReason) {
  console.log(response.filteredReason);
}

if (response.images.length == 0) {
  throw new Error("No images in the response.")
}

const image = response.images[0];

之后


import { initializeApp } from "firebase/app";
import {
  getAI,
  getGenerativeModel,
  GoogleAIBackend,
  ResponseModality,
  ImageConfigAspectRatio,
  FinishReason
} from "firebase/ai";

// TODO(developer): Replace the following with your app's Firebase configuration
const firebaseConfig = {
  // ...
};

// Initialize FirebaseApp
const firebaseApp = initializeApp(firebaseConfig);

// Initialize the Gemini Developer API backend service.
const ai = getAI(firebaseApp, { backend: new GoogleAIBackend() });

// Create a `GenerativeModel` instance with a model that supports your use case.
const model = getGenerativeModel(ai, {
  model: "GEMINI_NANOBANANA_MODEL_NAME",
  generationConfig: {
    responseModalities: [ResponseModality.IMAGE],
    imageConfig: {
      aspectRatio: ImageConfigAspectRatio.LANDSCAPE_4x3
    }
  },
});

// Provide an image generation prompt.
const prompt = "An astronaut riding a horse.";

// To generate an image, call `generateContent` with the text prompt.
const result = await model.generateContent(prompt);

// Handle the generated image.
try {
  const response = result.response;
  if (response.candidates?.[0].finishReason == FinishReason.NO_IMAGE) {
    // Handle the case where no images were generated.
  }
  const inlineDataParts = response.inlineDataParts();
  if (inlineDataParts?.[0]) {
    const image = inlineDataParts[0].inlineData;
    // Use this mimeType and base64 data to display the image using your preferred tooling.
    console.log(image.mimeType, image.data);
  }
} catch (err) {
  console.error('Prompt or candidate was blocked:', err);
}

Dart

之前


import 'package:firebase_ai/firebase_ai.dart';
import 'package:firebase_core/firebase_core.dart';
import 'firebase_options.dart';

// Initialize FirebaseApp
await Firebase.initializeApp(
  options: DefaultFirebaseOptions.currentPlatform,
);

// Initialize the Gemini Developer API backend service.
final ai = FirebaseAI.googleAI();

// Create an `ImagenModel` instance with an Imagen model that supports your use case.
final model = ai.imagenModel(model: 'IMAGEN_MODEL_NAME');

// Provide an image generation prompt.
const prompt = 'An astronaut riding a horse.';

// To generate an image, call `generateImages` with the text prompt.
final response = await model.generateImages(prompt);

if (response.images.isNotEmpty) {
  final image = response.images[0];
  // Process the image.
} else {
  // Handle the case where no images were generated.
  print('Error: No images were generated.');
}

之后


import 'package:firebase_ai/firebase_ai.dart';
import 'package:firebase_core/firebase_core.dart';
import 'firebase_options.dart';

// Initialize FirebaseApp
await Firebase.initializeApp(
  options: DefaultFirebaseOptions.currentPlatform,
);

// Initialize the Gemini Developer API backend service.
final ai = FirebaseAI.googleAI();

// Create a `GenerativeModel` instance with a Gemini model that supports image output.
final model = ai.generativeModel(
  model: 'GEMINI_NANOBANANA_MODEL_NAME',
  generationConfig: GenerationConfig(
    responseModalities: [ResponseModalities.image],
    imageConfig: ImageConfig(aspectRatio: ImageAspectRatio.landscape4x3)
  ),
);

// Provide a text prompt instructing the model to generate an image.
final prompt = [Content.text('An astronaut riding a horse.')];

// To generate an image, call `generateContent` with the text prompt.
final response = await model.generateContent(prompt);
if (response.inlineDataParts.isNotEmpty) {
  final imageBytes = response.inlineDataParts.first.bytes;
  // Process the image.
} else {
  // Handle the case where no images were generated.
  print('Error: No images were generated.');
}

Unity

之前


using Firebase.AI;

// Initialize the Gemini Developer API backend service
var ai = FirebaseAI.GetInstance(FirebaseAI.Backend.GoogleAI());

// Create an `ImagenModel` instance with a model that supports your use case
var model = ai.GetImagenModel(modelName: "IMAGEN_MODEL_NAME");

// Provide an image generation prompt
var prompt = "An astronaut riding a horse";

// To generate an image, call `generateImages` with the text prompt
var response = await model.GenerateImagesAsync(prompt: prompt);

// Handle the generated image
if (response.Images.Count == 0) {
  throw new Exception("No image in the response.");
}
var image = response.Images[0].AsTexture2D();

之后


using Firebase;
using Firebase.AI;

// Initialize the Gemini Developer API backend service.
var ai = FirebaseAI.GetInstance(FirebaseAI.Backend.GoogleAI());

// Create a `GenerativeModel` instance with a Gemini model that supports image output.
var model = ai.GetGenerativeModel(
  modelName: "GEMINI_NANOBANANA_MODEL_NAME",
  generationConfig: new GenerationConfig(
    responseModalities: new[] { ResponseModality.Image },
    imageConfig: new ImageConfig(aspectRatio: ImageConfig.AspectRatio.Landscape4x3)
  )
);

// Provide an image generation prompt.
var prompt = "An astronaut riding a horse";

// To generate an image, call `GenerateContentAsync` with the text prompt.
var response = await model.GenerateContentAsync(prompt);

if (response.Candidates.First().FinishReason == FinishReason.NoImage) {
  // Handle the case where no images were generated.
}

// Handle the generated image.
var imageParts = response.Candidates.First().Content.Parts
                         .OfType<ModelContent.InlineDataPart>()
                         .Where(part => part.MimeType == "image/png");

foreach (var imagePart in imageParts) {
  // Load the Image into a Unity Texture2D object.
  UnityEngine.Texture2D texture2D = new(2, 2);
  if (texture2D.LoadImage(imagePart.Data.ToArray())) {
    // Do something with the image.
  }
}

替换配置选项

本部分介绍了各种模型配置选项的替换选项，以帮助控制模型的回答。

安全设置

您可以使用 ImagenSafetySettings 为 Imagen 模型配置安全设置。不过，对于 Nano Banana 模型，您需要迁移到使用 SafetySetting。

模型配置参数

您可以使用 ImagenGenerationConfig 配置 Imagen 模型。不过，对于 Nano Banana 模型，您需要迁移到使用 GenerationConfig，并可选择使用嵌套的 ImageConfig（自 2026 年 5 月上旬发布的 SDK 版本开始提供）。

在 GenerationConfig 中，将回答模态设置为 IMAGE（如本指南前面的“之后”代码示例所示）。请注意，您可以选择将 Nano Banana 模型配置为同时返回 IMAGE 和 TEXT。

下表可帮助您了解如何将模型配置参数从 Imagen 模型迁移到 Nano Banana 模型：

Imagen 个模型 Gemini Nano Banana 个模型
addWatermark

不支持

Nano Banana 模型始终会返回带有 SynthID 水印的生成式图像。

aspectRatio

在 ImageConfig 中使用 aspectRatio

如需查看代码示例和支持的值，请参阅 Nano Banana 模型指南中的配置图片生成。

imageFormat

不支持

Nano Banana 模型始终以 PNG 格式返回生成的图片。

negativePrompt

不支持

请注意，反向提示是一项旧版功能，自 imagen-3.0-generate-002 起（或任何 Imagen 4 模型）便不再受支持。

numberOfImages

不支持

Nano Banana 模型始终会返回一张生成的图片。
作为权宜解决方法，您可以循环运行生成操作，以实现相同的结果。请注意，候选数量不能用作替代。

personGeneration

不支持

默认情况下，Nano Banana 模型允许生成人物图片。