Nano Banana 圖像生成功能
- 試用 Nano Banana 2 應用程式
- 你也可以根據提示詞自行建立：
-
-
採用 Nano Banana 2 技術生成 -
由 Nano Banana Pro 生成 -
採用 Nano Banana 2 技術生成 -
由 Nano Banana Pro 生成 -
由 Nano Banana Pro 生成 -
由 Nano Banana Pro 生成 -
由 Nano Banana Pro 生成 提示：「代表可愛小狗的圖示。背景為白色。以色彩豐富的觸覺 3D 風格製作圖示。沒有文字。"在 AI Studio 中使用 Nano Banana 製作圖示、貼圖和素材資源 -
採用 Nano Banana 2 技術生成
Nano Banana 是 Gemini 原生圖像生成功能的名稱。 Gemini 可透過對話互動生成及處理圖像， 並支援文字、圖像、影片或圖文組合。自由創作、編輯和反覆調整視覺內容，享有前所未有的掌控力。
Nano Banana 是指 Gemini API 中提供的三種不同模型：
- Nano Banana 2：Gemini 3.1 Flash Image 模型 (
gemini-3.1-flash-image)。這個模型是 Gemini 3 Pro Image 的高效能版本，經過最佳化，可滿足開發人員對速度和大量使用的需求。 - Nano Banana Pro：Gemini 3 Pro Image 模型 (
gemini-3-pro-image)。這個模型專為專業資產製作而設計，會運用進階推論 (「思考」) 功能，遵循複雜的指令並算繪高保真文字。 - Nano Banana：Gemini 2.5 Flash Image 模型 (
gemini-2.5-flash-image)。這個模型專為速度和效率而設計，適合處理大量低延遲的工作。
所有生成的圖片都會加上 SynthID 浮水印。
生成圖像 (文字轉圖像)
Python
from google import genai
from google.genai import types
from PIL import Image
client = genai.Client()
prompt = ("Create a picture of a nano banana dish in a fancy restaurant with a Gemini theme")
response = client.models.generate_content(
model="gemini-3.1-flash-image",
contents=[prompt],
)
for part in response.parts:
if part.text is not None:
print(part.text)
elif part.inline_data is not None:
image = part.as_image()
image.save("generated_image.png")
JavaScript
import { GoogleGenAI } from "@google/genai";
import * as fs from "node:fs";
async function main() {
const ai = new GoogleGenAI({});
const prompt =
"Create a picture of a nano banana dish in a fancy restaurant with a Gemini theme";
const response = await ai.models.generateContent({
model: "gemini-3.1-flash-image",
contents: prompt,
});
for (const part of response.candidates[0].content.parts) {
if (part.text) {
console.log(part.text);
} else if (part.inlineData) {
const imageData = part.inlineData.data;
const buffer = Buffer.from(imageData, "base64");
fs.writeFileSync("gemini-native-image.png", buffer);
console.log("Image saved as gemini-native-image.png");
}
}
}
main();
Go
package main
import (
"context"
"fmt"
"log"
"os"
"google.golang.org/genai"
)
func main() {
ctx := context.Background()
client, err := genai.NewClient(ctx, nil)
if err != nil {
log.Fatal(err)
}
result, _ := client.Models.GenerateContent(
ctx,
"gemini-3.1-flash-image",
genai.Text("Create a picture of a nano banana dish in a " +
" fancy restaurant with a Gemini theme"),
)
for _, part := range result.Candidates[0].Content.Parts {
if part.Text != "" {
fmt.Println(part.Text)
} else if part.InlineData != nil {
imageBytes := part.InlineData.Data
outputFilename := "gemini_generated_image.png"
_ = os.WriteFile(outputFilename, imageBytes, 0644)
}
}
}
Java
import com.google.genai.Client;
import com.google.genai.types.GenerateContentConfig;
import com.google.genai.types.GenerateContentResponse;
import com.google.genai.types.Part;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Paths;
public class TextToImage {
public static void main(String[] args) throws IOException {
try (Client client = new Client()) {
GenerateContentConfig config = GenerateContentConfig.builder()
.responseModalities("TEXT", "IMAGE")
.build();
GenerateContentResponse response = client.models.generateContent(
"gemini-3.1-flash-image",
"Create a picture of a nano banana dish in a fancy restaurant with a Gemini theme",
config);
for (Part part : response.parts()) {
if (part.text().isPresent()) {
System.out.println(part.text().get());
} else if (part.inlineData().isPresent()) {
var blob = part.inlineData().get();
if (blob.data().isPresent()) {
Files.write(Paths.get("_01_generated_image.png"), blob.data().get());
}
}
}
}
}
}
C#
using Google.GenAI;
using System;
using System.Collections.Generic;
using System.IO;
using System.Threading.Tasks;
public class TextToImage {
public static async Task Main(string[] args) {
var client = new Client();
var response = await client.Models.GenerateContentAsync(
model: "gemini-3.1-flash-image",
contents: new List<Part>
{
new Part { Text = "Create a picture of a nano banana dish in a fancy restaurant with a Gemini theme" }
}
);
foreach (var candidate in response.Candidates) {
foreach (var part in candidate.Content.Parts) {
if (part.Text != null) {
Console.WriteLine(part.Text);
} else if (part.InlineData != null) {
var imageBytes = Convert.FromBase64String(part.InlineData.Data);
await File.WriteAllBytesAsync("generated_image.png", imageBytes);
Console.WriteLine("Image saved as generated_image.png");
}
}
}
}
}
REST
curl -s -X POST \
"https://generativelanguage.googleapis.com/v1/models/gemini-3.1-flash-image:generateContent" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"contents": [{
"parts": [
{"text": "Create a picture of a nano banana dish in a fancy restaurant with a Gemini theme"}
]
}]
}'
圖像編輯 (文字和圖像轉圖像)
提醒：請確認您具備必要權限，可使用上傳的所有圖像。 請勿生成侵犯他人權利的內容，包括欺騙、騷擾或傷害他人的影片或圖像。使用這項生成式 AI 服務時，須遵守《使用限制政策》。
提供圖片並使用文字提示新增、移除或修改元素、變更風格，或調整色彩分級。
以下範例說明如何上傳 base64 編碼的圖片。
如要瞭解多張圖片、較大的酬載和支援的 MIME 類型，請參閱「圖片理解」頁面。
Python
from google import genai
from google.genai import types
from PIL import Image
client = genai.Client()
prompt = (
"Create a picture of my cat eating a nano-banana in a "
"fancy restaurant under the Gemini constellation",
)
image = Image.open("/path/to/cat_image.png")
response = client.models.generate_content(
model="gemini-3.1-flash-image",
contents=[prompt, image],
)
for part in response.parts:
if part.text is not None:
print(part.text)
elif part.inline_data is not None:
image = part.as_image()
image.save("generated_image.png")
JavaScript
import { GoogleGenAI } from "@google/genai";
import * as fs from "node:fs";
async function main() {
const ai = new GoogleGenAI({});
const imagePath = "path/to/cat_image.png";
const imageData = fs.readFileSync(imagePath);
const base64Image = imageData.toString("base64");
const prompt = [
{ text: "Create a picture of my cat eating a nano-banana in a" +
"fancy restaurant under the Gemini constellation" },
{
inlineData: {
mimeType: "image/png",
data: base64Image,
},
},
];
const response = await ai.models.generateContent({
model: "gemini-3.1-flash-image",
contents: prompt,
});
for (const part of response.candidates[0].content.parts) {
if (part.text) {
console.log(part.text);
} else if (part.inlineData) {
const imageData = part.inlineData.data;
const buffer = Buffer.from(imageData, "base64");
fs.writeFileSync("gemini-native-image.png", buffer);
console.log("Image saved as gemini-native-image.png");
}
}
}
main();
Go
package main
import (
"context"
"fmt"
"log"
"os"
"google.golang.org/genai"
)
func main() {
ctx := context.Background()
client, err := genai.NewClient(ctx, nil)
if err != nil {
log.Fatal(err)
}
imagePath := "/path/to/cat_image.png"
imgData, _ := os.ReadFile(imagePath)
parts := []*genai.Part{
genai.NewPartFromText("Create a picture of my cat eating a nano-banana in a fancy restaurant under the Gemini constellation"),
&genai.Part{
InlineData: &genai.Blob{
MIMEType: "image/png",
Data: imgData,
},
},
}
contents := []*genai.Content{
genai.NewContentFromParts(parts, genai.RoleUser),
}
result, _ := client.Models.GenerateContent(
ctx,
"gemini-3.1-flash-image",
contents,
)
for _, part := range result.Candidates[0].Content.Parts {
if part.Text != "" {
fmt.Println(part.Text)
} else if part.InlineData != nil {
imageBytes := part.InlineData.Data
outputFilename := "gemini_generated_image.png"
_ = os.WriteFile(outputFilename, imageBytes, 0644)
}
}
}
Java
import com.google.genai.Client;
import com.google.genai.types.Content;
import com.google.genai.types.GenerateContentConfig;
import com.google.genai.types.GenerateContentResponse;
import com.google.genai.types.Part;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Path;
import java.nio.file.Paths;
public class TextAndImageToImage {
public static void main(String[] args) throws IOException {
try (Client client = new Client()) {
GenerateContentConfig config = GenerateContentConfig.builder()
.responseModalities("TEXT", "IMAGE")
.build();
GenerateContentResponse response = client.models.generateContent(
"gemini-3.1-flash-image",
Content.fromParts(
Part.fromText("""
Create a picture of my cat eating a nano-banana in
a fancy restaurant under the Gemini constellation
"""),
Part.fromBytes(
Files.readAllBytes(
Path.of("src/main/resources/cat.jpg")),
"image/jpeg")),
config);
for (Part part : response.parts()) {
if (part.text().isPresent()) {
System.out.println(part.text().get());
} else if (part.inlineData().isPresent()) {
var blob = part.inlineData().get();
if (blob.data().isPresent()) {
Files.write(Paths.get("gemini_generated_image.png"), blob.data().get());
}
}
}
}
}
}
C#
using Google.GenAI;
using System;
using System.Collections.Generic;
using System.IO;
using System.Threading.Tasks;
public class TextAndImageToImage {
public static async Task Main(string[] args) {
var client = new Client();
var response = await client.Models.GenerateContentAsync(
model: "gemini-3.1-flash-image",
contents: new List<Part>
{
new Part { Text = "Create a picture of my cat eating a nano-banana in a fancy restaurant under the Gemini constellation" },
new Part
{
FileData = new FileData { FileUri = "file:///path/to/cat_image.png" }
}
}
);
foreach (var candidate in response.Candidates) {
foreach (var part in candidate.Content.Parts) {
if (part.Text != null) {
Console.WriteLine(part.Text);
} else if (part.InlineData != null) {
var imageBytes = Convert.FromBase64String(part.InlineData.Data);
await File.WriteAllBytesAsync("gemini_generated_image.png", imageBytes);
Console.WriteLine("Image saved as gemini_generated_image.png");
}
}
}
}
}
REST
curl -s -X POST \
"https://generativelanguage.googleapis.com/v1/models/gemini-3.1-flash-image:generateContent" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H 'Content-Type: application/json' \
-d "{
\"contents\": [{
\"parts\":[
{\"text\": \"'Create a picture of my cat eating a nano-banana in a fancy restaurant under the Gemini constellation\"},
{
\"inline_data\": {
\"mime_type\":\"image/jpeg\",
\"data\": \"<BASE64_IMAGE_DATA>\"
}
}
]
}]
}"
多輪圖像編輯
繼續透過對話生成及編輯圖像。建議使用對話或多輪對話功能，反覆編輯圖像。以下範例顯示生成光合作用資訊圖的提示。
Python
from google import genai
from google.genai import types
client = genai.Client()
chat = client.chats.create(
model="gemini-3.1-flash-image",
config=types.GenerateContentConfig(
response_modalities=['TEXT', 'IMAGE'],
tools=[{"google_search": {}}]
)
)
message = "Create a vibrant infographic that explains photosynthesis as if it were a recipe for a plant's favorite food. Show the \"ingredients\" (sunlight, water, CO2) and the \"finished dish\" (sugar/energy). The style should be like a page from a colorful kids' cookbook, suitable for a 4th grader."
response = chat.send_message(message)
for part in response.parts:
if part.text is not None:
print(part.text)
elif image:= part.as_image():
image.save("photosynthesis.png")
JavaScript
import { GoogleGenAI } from "@google/genai";
const ai = new GoogleGenAI({});
async function main() {
const chat = ai.chats.create({
model: "gemini-3.1-flash-image",
config: {
responseModalities: ['TEXT', 'IMAGE'],
tools: [{googleSearch: {}}],
},
});
}
await main();
const message = "Create a vibrant infographic that explains photosynthesis as if it were a recipe for a plant's favorite food. Show the \"ingredients\" (sunlight, water, CO2) and the \"finished dish\" (sugar/energy). The style should be like a page from a colorful kids' cookbook, suitable for a 4th grader."
let response = await chat.sendMessage({message});
for (const part of response.candidates[0].content.parts) {
if (part.text) {
console.log(part.text);
} else if (part.inlineData) {
const imageData = part.inlineData.data;
const buffer = Buffer.from(imageData, "base64");
fs.writeFileSync("photosynthesis.png", buffer);
console.log("Image saved as photosynthesis.png");
}
}
Go
package main
import (
"context"
"fmt"
"log"
"os"
"google.golang.org/genai"
)
func main() {
ctx := context.Background()
client, err := genai.NewClient(ctx, nil)
if err != nil {
log.Fatal(err)
}
defer client.Close()
model := client.GenerativeModel("gemini-3.1-flash-image")
model.GenerationConfig = &pb.GenerationConfig{
ResponseModalities: []pb.ResponseModality{genai.Text, genai.Image},
}
chat := model.StartChat()
message := "Create a vibrant infographic that explains photosynthesis as if it were a recipe for a plant's favorite food. Show the \"ingredients\" (sunlight, water, CO2) and the \"finished dish\" (sugar/energy). The style should be like a page from a colorful kids' cookbook, suitable for a 4th grader."
resp, err := chat.SendMessage(ctx, genai.Text(message))
if err != nil {
log.Fatal(err)
}
for _, part := range resp.Candidates[0].Content.Parts {
if txt, ok := part.(genai.Text); ok {
fmt.Printf("%s", string(txt))
} else if img, ok := part.(genai.ImageData); ok {
err := os.WriteFile("photosynthesis.png", img.Data, 0644)
if err != nil {
log.Fatal(err)
}
}
}
}
Java
import com.google.genai.Chat;
import com.google.genai.Client;
import com.google.genai.types.Content;
import com.google.genai.types.GenerateContentConfig;
import com.google.genai.types.GenerateContentResponse;
import com.google.genai.types.GoogleSearch;
import com.google.genai.types.ImageConfig;
import com.google.genai.types.Part;
import com.google.genai.types.RetrievalConfig;
import com.google.genai.types.Tool;
import com.google.genai.types.ToolConfig;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Path;
import java.nio.file.Paths;
public class MultiturnImageEditing {
public static void main(String[] args) throws IOException {
try (Client client = new Client()) {
GenerateContentConfig config = GenerateContentConfig.builder()
.responseModalities("TEXT", "IMAGE")
.tools(Tool.builder()
.googleSearch(GoogleSearch.builder().build())
.build())
.build();
Chat chat = client.chats.create("gemini-3.1-flash-image", config);
GenerateContentResponse response = chat.sendMessage("""
Create a vibrant infographic that explains photosynthesis
as if it were a recipe for a plant's favorite food.
Show the "ingredients" (sunlight, water, CO2)
and the "finished dish" (sugar/energy).
The style should be like a page from a colorful
kids' cookbook, suitable for a 4th grader.
""");
for (Part part : response.parts()) {
if (part.text().isPresent()) {
System.out.println(part.text().get());
} else if (part.inlineData().isPresent()) {
var blob = part.inlineData().get();
if (blob.data().isPresent()) {
Files.write(Paths.get("photosynthesis.png"), blob.data().get());
}
}
}
// ...
}
}
}
C#
using Google.GenAI;
using System;
using System.Collections.Generic;
using System.IO;
using System.Threading.Tasks;
public class MultiturnImageEditing {
public static async Task Main(string[] args) {
var client = new Client();
var response = await client.Models.GenerateContentAsync(
model: "gemini-3.1-flash-image",
contents: new List<Part>
{
new Part { Text = "Create a vibrant infographic that explains photosynthesis as if it were a recipe for a plant's favorite food. Show the \"ingredients\" (sunlight, water, CO2) and the \"finished dish\" (sugar/energy). The style should be like a page from a colorful kids' cookbook, suitable for a 4th grader." }
},
config: new GenerateContentConfig
{
ResponseModalities = new List<string> { "TEXT", "IMAGE" },
Tools = new List<Tool> { new Tool { GoogleSearch = new GoogleSearch() } }
}
);
foreach (var candidate in response.Candidates) {
foreach (var part in candidate.Content.Parts) {
if (part.Text != null) {
Console.WriteLine(part.Text);
} else if (part.InlineData != null) {
var imageBytes = Convert.FromBase64String(part.InlineData.Data);
await File.WriteAllBytesAsync("photosynthesis.png", imageBytes);
Console.WriteLine("Image saved as photosynthesis.png");
}
}
}
}
}
REST
curl -s -X POST \
"https://generativelanguage.googleapis.com/v1/models/gemini-3.1-flash-image:generateContent" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"contents": [{
"role": "user",
"parts": [
{"text": "Create a vibrant infographic that explains photosynthesis as if it were a recipe for a plants favorite food. Show the \"ingredients\" (sunlight, water, CO2) and the \"finished dish\" (sugar/energy). The style should be like a page from a colorful kids cookbook, suitable for a 4th grader."}
]
}],
"generationConfig": {
"responseModalities": ["TEXT", "IMAGE"]
}
}'
接著，您可以在同一個對話中，要求將圖片上的文字改為西班牙文。
Python
message = "Update this infographic to be in Spanish. Do not change any other elements of the image."
aspect_ratio = "16:9" # "1:1","1:4","1:8","2:3","3:2","3:4","4:1","4:3","4:5","5:4","8:1","9:16","16:9","21:9"
resolution = "2K" # "512", "1K", "2K", "4K"
response = chat.send_message(message,
config=types.GenerateContentConfig(
response_format={"image": {aspect_ratio: aspect_ratio, image_size: resolution}},
))
for part in response.parts:
if part.text is not None:
print(part.text)
elif image:= part.as_image():
image.save("photosynthesis_spanish.png")
JavaScript
const message = 'Update this infographic to be in Spanish. Do not change any other elements of the image.';
const aspectRatio = '16:9';
const resolution = '2K';
let response = await chat.sendMessage({
message,
config: {
responseModalities: ['TEXT', 'IMAGE'],
responseFormat: {
image: {
aspectRatio: aspectRatio,
imageSize: resolution,
}
},
tools: [{googleSearch: {}}],
},
});
for (const part of response.candidates[0].content.parts) {
if (part.text) {
console.log(part.text);
} else if (part.inlineData) {
const imageData = part.inlineData.data;
const buffer = Buffer.from(imageData, "base64");
fs.writeFileSync("photosynthesis2.png", buffer);
console.log("Image saved as photosynthesis2.png");
}
}
Go
message = "Update this infographic to be in Spanish. Do not change any other elements of the image."
aspect_ratio = "16:9" // "1:1","1:4","1:8","2:3","3:2","3:4","4:1","4:3","4:5","5:4","8:1","9:16","16:9","21:9"
resolution = "2K" // "512", "1K", "2K", "4K"
model.GenerationConfig.ImageConfig = &pb.ImageConfig{
AspectRatio: aspect_ratio,
ImageSize: resolution,
}
resp, err = chat.SendMessage(ctx, genai.Text(message))
if err != nil {
log.Fatal(err)
}
for _, part := range resp.Candidates[0].Content.Parts {
if txt, ok := part.(genai.Text); ok {
fmt.Printf("%s", string(txt))
} else if img, ok := part.(genai.ImageData); ok {
err := os.WriteFile("photosynthesis_spanish.png", img.Data, 0644)
if err != nil {
log.Fatal(err)
}
}
}
Java
String aspectRatio = "16:9"; // "1:1","1:4","1:8","2:3","3:2","3:4","4:1","4:3","4:5","5:4","8:1","9:16","16:9","21:9"
String resolution = "2K"; // "512", "1K", "2K", "4K"
config = GenerateContentConfig.builder()
.responseModalities("TEXT", "IMAGE")
.imageConfig(ImageConfig.builder()
.aspectRatio(aspectRatio)
.imageSize(resolution)
.build())
.build();
response = chat.sendMessage(
"Update this infographic to be in Spanish. " +
"Do not change any other elements of the image.",
config);
for (Part part : response.parts()) {
if (part.text().isPresent()) {
System.out.println(part.text().get());
} else if (part.inlineData().isPresent()) {
var blob = part.inlineData().get();
if (blob.data().isPresent()) {
Files.write(Paths.get("photosynthesis_spanish.png"), blob.data().get());
}
}
}
C#
using Google.GenAI;
using System;
using System.Collections.Generic;
using System.IO;
using System.Threading.Tasks;
public class MultiturnImageEditingSpanish {
public static async Task Main(string[] args) {
var client = new Client();
var response = await client.Models.GenerateContentAsync(
model: "gemini-3.1-flash-image",
contents: new List<Part>
{
new Part { Text = "Update this infographic to be in Spanish. Do not change any other elements of the image." }
},
config: new GenerateContentConfig
{
ResponseModalities = new List<string> { "TEXT", "IMAGE" },
ImageConfig = new ImageConfig
{
AspectRatio = "16:9",
ImageSize = "2K"
}
}
);
foreach (var candidate in response.Candidates) {
foreach (var part in candidate.Content.Parts) {
if (part.Text != null) {
Console.WriteLine(part.Text);
} else if (part.InlineData != null) {
var imageBytes = Convert.FromBase64String(part.InlineData.Data);
await File.WriteAllBytesAsync("photosynthesis_spanish.png", imageBytes);
Console.WriteLine("Image saved as photosynthesis_spanish.png");
}
}
}
}
}
REST
curl -s -X POST \
"https://generativelanguage.googleapis.com/v1/models/gemini-3.1-flash-image:generateContent" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H 'Content-Type: application/json' \
-d '{
"contents": [
{
"role": "user",
"parts": [{"text": "Create a vibrant infographic that explains photosynthesis..."}]
},
{
"role": "model",
"parts": [{"inline_data": {"mime_type": "image/png", "data": "<PREVIOUS_IMAGE_DATA>"}}]
},
{
"role": "user",
"parts": [{"text": "Update this infographic to be in Spanish. Do not change any other elements of the image."}]
}
],
"tools": [{"google_search": {}}],
"generationConfig": {
"responseModalities": ["TEXT", "IMAGE"],
"responseFormat": {
"image": {
"aspectRatio": "16:9",
"imageSize": "2K"
}
}
}
}'
全新 Gemini 3 Image 模型
Gemini 3 提供最先進的圖像生成和編輯模型。Gemini 3.1 Flash Image 經過最佳化處理，速度快且適合大量使用情境；Gemini 3 Pro Image 則經過最佳化處理，適合製作專業素材。這類模型具備進階推論能力，可處理最困難的工作流程，擅長執行複雜的多輪建立和修改任務。
- 高解析度輸出：內建生成 1K、2K 和 4K 圖像的功能。
- Gemini 3.1 Flash Image 新增較小的 512 (0.5K) 解析度。
- 進階文字算繪：可為資訊圖表、菜單、圖表和行銷素材資源生成清晰的風格化文字。
- 以 Google 搜尋做為基準：模型可使用 Google 搜尋做為工具，根據即時資料驗證事實並生成圖像 (例如目前的天氣地圖、股票圖表、近期事件)。
- Gemini 3.1 Flash Image 除了整合 Google 網頁搜尋，也整合了 Google 圖片搜尋，可做為生成圖像的基準。
- 思考模式：模型會運用「思考」程序，推論複雜的提示。這項功能會生成臨時的「想法圖像」(顯示在後端，但不會收費)，以改善構圖，然後再製作最終的高品質輸出內容。
- 最多 14 張參考圖片：現在最多可混合 14 張參考圖片，生成最終圖片。
- 新增顯示比例：Gemini 3.1 Flash Image 新增 1:4、4:1、1:8 和 8:1 顯示比例。
最多可使用 14 張參考圖片
Gemini 3 圖像模型最多可混合 14 張參考圖片。這 14 張圖片可包括：
| Gemini 3.1 Flash Image | Gemini 3 Pro Image |
|---|---|
| 最多 10 張高保真物件圖片，可納入最終圖片 | 最多 6 張高保真物件圖片，可加入最終圖片 |
| 最多 4 張角色圖片，確保角色一致性 | 最多 5 張角色圖片，確保角色一致性 |
| 不適用 | 最多可上傳 3 張圖片做為風格參考 |
Python
from google import genai
from google.genai import types
from PIL import Image
prompt = "An office group photo of these people, they are making funny faces."
aspect_ratio = "5:4" # "1:1","1:4","1:8","2:3","3:2","3:4","4:1","4:3","4:5","5:4","8:1","9:16","16:9","21:9"
resolution = "2K" # "512", "1K", "2K", "4K"
client = genai.Client()
response = client.models.generate_content(
model="gemini-3.1-flash-image",
contents=[
prompt,
Image.open('person1.png'),
Image.open('person2.png'),
Image.open('person3.png'),
Image.open('person4.png'),
Image.open('person5.png'),
],
config=types.GenerateContentConfig(
response_modalities=['TEXT', 'IMAGE'],
response_format={"image": {aspect_ratio: aspect_ratio, image_size: resolution}},
)
)
for part in response.parts:
if part.text is not None:
print(part.text)
elif image:= part.as_image():
image.save("office.png")
JavaScript
import { GoogleGenAI } from "@google/genai";
import * as fs from "node:fs";
async function main() {
const ai = new GoogleGenAI({});
const prompt =
'An office group photo of these people, they are making funny faces.';
const aspectRatio = '5:4';
const resolution = '2K';
const contents = [
{ text: prompt },
{
inlineData: {
mimeType: "image/jpeg",
data: base64ImageFile1,
},
},
{
inlineData: {
mimeType: "image/jpeg",
data: base64ImageFile2,
},
},
{
inlineData: {
mimeType: "image/jpeg",
data: base64ImageFile3,
},
},
{
inlineData: {
mimeType: "image/jpeg",
data: base64ImageFile4,
},
},
{
inlineData: {
mimeType: "image/jpeg",
data: base64ImageFile5,
},
}
];
const response = await ai.models.generateContent({
model: 'gemini-3.1-flash-image',
contents: contents,
config: {
responseModalities: ['TEXT', 'IMAGE'],
responseFormat: {
image: {
aspectRatio: aspectRatio,
imageSize: resolution,
}
},
},
});
for (const part of response.candidates[0].content.parts) {
if (part.text) {
console.log(part.text);
} else if (part.inlineData) {
const imageData = part.inlineData.data;
const buffer = Buffer.from(imageData, "base64");
fs.writeFileSync("image.png", buffer);
console.log("Image saved as image.png");
}
}
}
main();
Go
package main
import (
"context"
"fmt"
"log"
"os"
"google.golang.org/genai"
)
func main() {
ctx := context.Background()
client, err := genai.NewClient(ctx, nil)
if err != nil {
log.Fatal(err)
}
defer client.Close()
model := client.GenerativeModel("gemini-3.1-flash-image")
model.GenerationConfig = &pb.GenerationConfig{
ResponseModalities: []pb.ResponseModality{genai.Text, genai.Image},
ImageConfig: &pb.ImageConfig{
AspectRatio: "5:4",
ImageSize: "2K",
},
}
img1, err := os.ReadFile("person1.png")
if err != nil { log.Fatal(err) }
img2, err := os.ReadFile("person2.png")
if err != nil { log.Fatal(err) }
img3, err := os.ReadFile("person3.png")
if err != nil { log.Fatal(err) }
img4, err := os.ReadFile("person4.png")
if err != nil { log.Fatal(err) }
img5, err := os.ReadFile("person5.png")
if err != nil { log.Fatal(err) }
parts := []genai.Part{
genai.Text("An office group photo of these people, they are making funny faces."),
genai.ImageData{MIMEType: "image/png", Data: img1},
genai.ImageData{MIMEType: "image/png", Data: img2},
genai.ImageData{MIMEType: "image/png", Data: img3},
genai.ImageData{MIMEType: "image/png", Data: img4},
genai.ImageData{MIMEType: "image/png", Data: img5},
}
resp, err := model.GenerateContent(ctx, parts...)
if err != nil {
log.Fatal(err)
}
for _, part := range resp.Candidates[0].Content.Parts {
if txt, ok := part.(genai.Text); ok {
fmt.Printf("%s", string(txt))
} else if img, ok := part.(genai.ImageData); ok {
err := os.WriteFile("office.png", img.Data, 0644)
if err != nil {
log.Fatal(err)
}
}
}
}
Java
import com.google.genai.Client;
import com.google.genai.types.Content;
import com.google.genai.types.GenerateContentConfig;
import com.google.genai.types.GenerateContentResponse;
import com.google.genai.types.ImageConfig;
import com.google.genai.types.Part;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Path;
import java.nio.file.Paths;
public class GroupPhoto {
public static void main(String[] args) throws IOException {
try (Client client = new Client()) {
GenerateContentConfig config = GenerateContentConfig.builder()
.responseModalities("TEXT", "IMAGE")
.imageConfig(ImageConfig.builder()
.aspectRatio("5:4")
.imageSize("2K")
.build())
.build();
GenerateContentResponse response = client.models.generateContent(
"gemini-3.1-flash-image",
Content.fromParts(
Part.fromText("An office group photo of these people, they are making funny faces."),
Part.fromBytes(Files.readAllBytes(Path.of("person1.png")), "image/png"),
Part.fromBytes(Files.readAllBytes(Path.of("person2.png")), "image/png"),
Part.fromBytes(Files.readAllBytes(Path.of("person3.png")), "image/png"),
Part.fromBytes(Files.readAllBytes(Path.of("person4.png")), "image/png"),
Part.fromBytes(Files.readAllBytes(Path.of("person5.png")), "image/png")
), config);
for (Part part : response.parts()) {
if (part.text().isPresent()) {
System.out.println(part.text().get());
} else if (part.inlineData().isPresent()) {
var blob = part.inlineData().get();
if (blob.data().isPresent()) {
Files.write(Paths.get("office.png"), blob.data().get());
}
}
}
}
}
}
C#
using Google.GenAI;
using System;
using System.Collections.Generic;
using System.IO;
using System.Threading.Tasks;
public class GroupPhoto {
public static async Task Main(string[] args) {
var client = new Client();
var response = await client.Models.GenerateContentAsync(
model: "gemini-3.1-flash-image",
contents: new List<Part>
{
new Part { Text = "An office group photo of these people, they are making funny faces." },
new Part { FileData = new FileData { FileUri = "file:///person1.png" } },
new Part { FileData = new FileData { FileUri = "file:///person2.png" } },
new Part { FileData = new FileData { FileUri = "file:///person3.png" } },
new Part { FileData = new FileData { FileUri = "file:///person4.png" } },
new Part { FileData = new FileData { FileUri = "file:///person5.png" } }
},
config: new GenerateContentConfig
{
ResponseModalities = new List<string> { "TEXT", "IMAGE" },
ImageConfig = new ImageConfig
{
AspectRatio = "5:4",
ImageSize = "2K"
}
}
);
foreach (var candidate in response.Candidates) {
foreach (var part in candidate.Content.Parts) {
if (part.Text != null) {
Console.WriteLine(part.Text);
} else if (part.InlineData != null) {
var imageBytes = Convert.FromBase64String(part.InlineData.Data);
await File.WriteAllBytesAsync("office.png", imageBytes);
Console.WriteLine("Image saved as office.png");
}
}
}
}
}
REST
curl -s -X POST \
"https://generativelanguage.googleapis.com/v1/models/gemini-3.1-flash-image:generateContent" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H 'Content-Type: application/json' \
-d "{
\"contents\": [{
\"parts\":[
{\"text\": \"An office group photo of these people, they are making funny faces.\"},
{\"inline_data\": {\"mime_type\":\"image/png\", \"data\": \"<BASE64_DATA_IMG_1>\"}},
{\"inline_data\": {\"mime_type\":\"image/png\", \"data\": \"<BASE64_DATA_IMG_2>\"}},
{\"inline_data\": {\"mime_type\":\"image/png\", \"data\": \"<BASE64_DATA_IMG_3>\"}},
{\"inline_data\": {\"mime_type\":\"image/png\", \"data\": \"<BASE64_DATA_IMG_4>\"}},
{\"inline_data\": {\"mime_type\":\"image/png\", \"data\": \"<BASE64_DATA_IMG_5>\"}}
]
}],
\"generationConfig\": {
\"responseModalities\": [\"TEXT\", \"IMAGE\"],
\"responseFormat\": {
\"image\": {
\"aspectRatio\": \"5:4\",
\"imageSize\": \"2K\"
}
}
}
}"
以 Google 搜尋建立基準
使用 Google 搜尋工具，根據天氣預報、股票圖表或近期活動等即時資訊生成圖片。
請注意，使用 Google 搜尋建立基準生成圖像時，系統不會將圖片搜尋結果傳送至生成模型，且不會顯示在回覆中 (請參閱以 Google 搜尋建立圖像基準)。
Python
from google import genai
prompt = "Visualize the current weather forecast for the next 5 days in San Francisco as a clean, modern weather chart. Add a visual on what I should wear each day"
aspect_ratio = "16:9" # "1:1","1:4","1:8","2:3","3:2","3:4","4:1","4:3","4:5","5:4","8:1","9:16","16:9","21:9"
client = genai.Client()
response = client.models.generate_content(
model="gemini-3.1-flash-image",
contents=prompt,
config=types.GenerateContentConfig(
response_modalities=['Text', 'Image'],
response_format={"image": {aspect_ratio: aspect_ratio,}},
tools=[{"google_search": {}}]
)
)
for part in response.parts:
if part.text is not None:
print(part.text)
elif image:= part.as_image():
image.save("weather.png")
JavaScript
import { GoogleGenAI } from "@google/genai";
import * as fs from "node:fs";
async function main() {
const ai = new GoogleGenAI({});
const prompt = 'Visualize the current weather forecast for the next 5 days in San Francisco as a clean, modern weather chart. Add a visual on what I should wear each day';
const aspectRatio = '16:9';
const resolution = '2K';
const response = await ai.models.generateContent({
model: 'gemini-3.1-flash-image',
contents: prompt,
config: {
responseModalities: ['TEXT', 'IMAGE'],
responseFormat: {
image: {
aspectRatio: aspectRatio,
imageSize: resolution,
}
},
tools: [{ googleSearch: {} }]
},
});
for (const part of response.candidates[0].content.parts) {
if (part.text) {
console.log(part.text);
} else if (part.inlineData) {
const imageData = part.inlineData.data;
const buffer = Buffer.from(imageData, "base64");
fs.writeFileSync("image.png", buffer);
console.log("Image saved as image.png");
}
}
}
main();
Java
import com.google.genai.Client;
import com.google.genai.types.GenerateContentConfig;
import com.google.genai.types.GenerateContentResponse;
import com.google.genai.types.GoogleSearch;
import com.google.genai.types.ImageConfig;
import com.google.genai.types.Part;
import com.google.genai.types.Tool;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Paths;
public class SearchGrounding {
public static void main(String[] args) throws IOException {
try (Client client = new Client()) {
GenerateContentConfig config = GenerateContentConfig.builder()
.responseModalities("TEXT", "IMAGE")
.imageConfig(ImageConfig.builder()
.aspectRatio("16:9")
.build())
.tools(Tool.builder()
.googleSearch(GoogleSearch.builder().build())
.build())
.build();
GenerateContentResponse response = client.models.generateContent(
"gemini-3.1-flash-image", """
Visualize the current weather forecast for the next 5 days
in San Francisco as a clean, modern weather chart.
Add a visual on what I should wear each day
""",
config);
for (Part part : response.parts()) {
if (part.text().isPresent()) {
System.out.println(part.text().get());
} else if (part.inlineData().isPresent()) {
var blob = part.inlineData().get();
if (blob.data().isPresent()) {
Files.write(Paths.get("weather.png"), blob.data().get());
}
}
}
}
}
}
C#
using Google.GenAI;
using System;
using System.Collections.Generic;
using System.IO;
using System.Threading.Tasks;
public class SearchGrounding {
public static async Task Main(string[] args) {
var client = new Client();
var response = await client.Models.GenerateContentAsync(
model: "gemini-3.1-flash-image",
contents: new List<Part>
{
new Part { Text = "Visualize the current weather forecast for the next 5 days in San Francisco as a clean, modern weather chart. Add a visual on what I should wear each day" }
},
config: new GenerateContentConfig
{
ResponseModalities = new List<string> { "TEXT", "IMAGE" },
ImageConfig = new ImageConfig
{
AspectRatio = "16:9"
},
Tools = new List<Tool> { new Tool { GoogleSearch = new GoogleSearch() } }
}
);
foreach (var candidate in response.Candidates) {
foreach (var part in candidate.Content.Parts) {
if (part.Text != null) {
Console.WriteLine(part.Text);
} else if (part.InlineData != null) {
var imageBytes = Convert.FromBase64String(part.InlineData.Data);
await File.WriteAllBytesAsync("weather.png", imageBytes);
Console.WriteLine("Image saved as weather.png");
}
}
}
}
}
REST
curl -s -X POST \
"https://generativelanguage.googleapis.com/v1/models/gemini-3.1-flash-image:generateContent" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"contents": [{"parts": [{"text": "Visualize the current weather forecast for the next 5 days in San Francisco as a clean, modern weather chart. Add a visual on what I should wear each day"}]}],
"tools": [{"google_search": {}}],
"generationConfig": {
"responseModalities": ["TEXT", "IMAGE"],
"responseFormat": {
"image": {"aspectRatio": "16:9"}
}
}
}'
回應會包含 groundingMetadata，其中含有下列必填欄位：
searchEntryPoint：包含 HTML 和 CSS，可轉譯必要的搜尋建議。groundingChunks：傳回用於生成圖片的前 3 個網路來源
以 Google 搜尋圖片功能建立基準 (3.1 Flash)
以 Google 搜尋強化事實基礎生成圖像時，模型會使用透過 Google 搜尋擷取的網路圖片做為視覺背景。圖片搜尋是現有「以 Google 搜尋強化事實基礎」工具中的新搜尋類型，可與標準的Google 網頁搜尋並用。
如要啟用圖片搜尋功能，請在 API 要求中設定 googleSearch 工具，並在 searchTypes 物件中指定 imageSearch。圖片搜尋可以單獨使用，也可以與網頁搜尋搭配使用。
請注意，圖片的 Google 搜尋基礎功能無法用於搜尋人物。
Python
from google import genai
prompt = "A detailed painting of a Timareta butterfly resting on a flower"
client = genai.Client()
response = client.models.generate_content(
model="gemini-3.1-flash-image",
contents=prompt,
config=types.GenerateContentConfig(
response_modalities=["IMAGE"],
tools=[
types.Tool(google_search=types.GoogleSearch(
search_types=types.SearchTypes(
web_search=types.WebSearch(),
image_search=types.ImageSearch()
)
))
]
)
)
# Display grounding sources if available
if response.candidates and response.candidates[0].grounding_metadata and response.candidates[0].grounding_metadata.search_entry_point:
display(HTML(response.candidates[0].grounding_metadata.search_entry_point.rendered_content))
JavaScript
import { GoogleGenAI } from "@google/genai";
async function main() {
const ai = new GoogleGenAI({});
const prompt = "A detailed painting of a Timareta butterfly resting on a flower";
const response = await ai.models.generateContent({
model: "gemini-3.1-flash-image",
contents: prompt,
config: {
responseModalities: ["IMAGE"],
tools: [
{
googleSearch: {
searchTypes: {
webSearch: {},
imageSearch: {}
}
}
}
]
}
});
// Display grounding sources if available
if (response.candidates && response.candidates[0].groundingMetadata && response.candidates[0].groundingMetadata.searchEntryPoint) {
console.log(response.candidates[0].groundingMetadata.searchEntryPoint.renderedContent);
}
}
main();
Go
package main
import (
"context"
"fmt"
"log"
"google.golang.org/genai"
pb "google.golang.org/genai/schema"
)
func main() {
ctx := context.Background()
client, err := genai.NewClient(ctx, nil)
if err != nil {
log.Fatal(err)
}
defer client.Close()
model := client.GenerativeModel("gemini-3.1-flash-image")
model.Tools = []*pb.Tool{
{
GoogleSearch: &pb.GoogleSearch{
SearchTypes: &pb.SearchTypes{
WebSearch: &pb.WebSearch{},
ImageSearch: &pb.ImageSearch{},
},
},
},
}
model.GenerationConfig = &pb.GenerationConfig{
ResponseModalities: []pb.ResponseModality{genai.Image},
}
prompt := "A detailed painting of a Timareta butterfly resting on a flower"
resp, err := model.GenerateContent(ctx, genai.Text(prompt))
if err != nil {
log.Fatal(err)
}
if resp.Candidates[0].GroundingMetadata != nil && resp.Candidates[0].GroundingMetadata.SearchEntryPoint != nil {
fmt.Println(resp.Candidates[0].GroundingMetadata.SearchEntryPoint.RenderedContent)
}
}
Java
import com.google.genai.Client;
import com.google.genai.types.GenerateContentConfig;
import com.google.genai.types.GenerateContentResponse;
import com.google.genai.types.GoogleSearch;
import com.google.genai.types.ImageSearch;
import com.google.genai.types.SearchTypes;
import com.google.genai.types.Tool;
import com.google.genai.types.WebSearch;
import java.io.IOException;
public class ImageSearchGrounding {
public static void main(String[] args) throws IOException {
try (Client client = new Client()) {
GenerateContentConfig config = GenerateContentConfig.builder()
.responseModalities("IMAGE")
.tools(Tool.builder()
.googleSearch(GoogleSearch.builder()
.searchTypes(SearchTypes.builder()
.webSearch(WebSearch.builder().build())
.imageSearch(ImageSearch.builder().build())
.build())
.build())
.build())
.build();
GenerateContentResponse response = client.models.generateContent(
"gemini-3.1-flash-image",
"A detailed painting of a Timareta butterfly resting on a flower",
config);
if (response.candidates().isPresent() && !response.candidates().get().isEmpty()) {
var candidate = response.candidates().get().get(0);
if (candidate.groundingMetadata().isPresent() && candidate.groundingMetadata().get().searchEntryPoint().isPresent()) {
System.out.println(candidate.groundingMetadata().get().searchEntryPoint().get().renderedContent().orElse(""));
}
}
}
}
}
C#
using Google.GenAI;
using System;
using System.Collections.Generic;
using System.Threading.Tasks;
public class ImageSearchGrounding {
public static async Task Main(string[] args) {
var client = new Client();
var response = await client.Models.GenerateContentAsync(
model: "gemini-3.1-flash-image",
contents: new List<Part>
{
new Part { Text = "A detailed painting of a Timareta butterfly resting on a flower" }
},
config: new GenerateContentConfig
{
ResponseModalities = new List<string> { "IMAGE" },
Tools = new List<Tool>
{
new Tool
{
GoogleSearch = new GoogleSearch
{
SearchTypes = new SearchTypes
{
WebSearch = new WebSearch(),
ImageSearch = new ImageSearch()
}
}
}
}
}
);
foreach (var candidate in response.Candidates) {
if (candidate.GroundingMetadata != null && candidate.GroundingMetadata.SearchEntryPoint != null) {
Console.WriteLine(candidate.GroundingMetadata.SearchEntryPoint.RenderedContent);
}
}
}
}
REST
curl -s -X POST \
"https://generativelanguage.googleapis.com/v1/models/gemini-3.1-flash-image:generateContent" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"contents": [{"parts": [{"text": "A detailed painting of a Timareta butterfly resting on a flower"}]}],
"tools": [{"google_search": {"searchTypes": {"webSearch": {}, "imageSearch": {}}}}],
"generationConfig": {
"responseModalities": ["IMAGE"]
}
}'
螢幕規定
在「以 Google 搜尋強化事實基礎」功能中使用圖片搜尋時，必須遵守下列條件：
- 來源出處：你必須提供含有來源圖片的網頁連結 (「含有頁面」，而非圖片檔案本身)，且使用者可辨識為連結。
- 直接導覽：如果選擇顯示來源圖片，必須提供從來源圖片到所含來源網頁的直接單點路徑。任何會延遲或阻礙使用者存取來源網頁的實作方式皆不允許，包括但不限於任何多重點擊路徑，或使用中繼圖片檢視器。
回應
如果是以圖片搜尋結果為基準的回覆，API 會提供清楚的歸因和中繼資料，將輸出內容連結至經過驗證的來源。groundingMetadata 物件中的主要欄位包括：
imageSearchQueries：模型用於視覺內容 (圖片搜尋) 的特定查詢。groundingChunks：包含所擷取結果的來源資訊。如果是圖片來源，系統會使用新的圖片區塊類型，以重新導向網址的形式傳回。這段內容包括：uri：用於歸因的網頁網址 (到達網頁)。image_uri：圖片的直接網址。
groundingSupports：提供具體對應，將生成的內容連結至相關的引用來源。searchEntryPoint：包含「Google 搜尋」資訊方塊，內含符合規定的 HTML 和 CSS，可顯示搜尋建議。
從影片生成圖片 (3.1 Flash)
影片轉圖像功能可使用影片內容做為多模態參考，生成新圖像。這項功能可製作高品質的影片縮圖、電影海報、摘要資訊圖表，或以影片場景為靈感的新作品。
生成圖片時，模型會分析影片影格的內容 (最多可達模型輸入權杖限制的 131,072 個權杖)，擷取視覺主題和重要事件，然後搭配文字提示合成輸出圖片。
您可以直接在 API 要求中傳遞公開的 YouTube 網址，也可以使用 Files API 上傳本機影片檔案。
Python
from google import genai
from google.genai import types
client = genai.Client()
# Pass a public YouTube video URL as part of the contents
response = client.models.generate_content(
model="gemini-3.1-flash-image",
contents=[
types.Part(
file_data=types.FileData(file_uri="https://www.youtube.com/watch?v=UTdfxFyOQTI"),
video_metadata=types.VideoMetadata(fps=0.5)
),
"Generate a poster image that captures the key themes of this video."
],
config=types.GenerateContentConfig(
response_modalities=["TEXT", "IMAGE"]
)
)
# Save the generated image part
for part in response.parts:
if part.inline_data is not None:
image = part.as_image()
image.save("video_poster.png")
print("Image saved as video_poster.png")
JavaScript
import { GoogleGenAI } from "@google/genai";
import * as fs from "node:fs";
async function main() {
const ai = new GoogleGenAI({});
const response = await ai.models.generateContent({
model: "gemini-3.1-flash-image",
contents: [
{
fileData: {
fileUri: "https://www.youtube.com/watch?v=UTdfxFyOQTI",
},
videoMetadata: {
fps: 0.5
}
},
{ text: "Generate a poster image that captures the key themes of this video." }
],
config: {
responseModalities: ["TEXT", "IMAGE"]
}
});
for (const part of response.candidates[0].content.parts) {
if (part.inlineData) {
const imageData = part.inlineData.data;
const buffer = Buffer.from(imageData, "base64");
fs.writeFileSync("video_poster.png", buffer);
console.log("Image saved as video_poster.png");
}
}
}
main();
Go
package main
import (
"context"
"log"
"os"
"google.golang.org/genai"
)
func main() {
ctx := context.Background()
client, err := genai.NewClient(ctx, nil)
if err != nil {
log.Fatal(err)
}
videoPart := genai.NewPartFromURI("https://www.youtube.com/watch?v=UTdfxFyOQTI", "video/mp4")
videoPart.VideoMetadata = &genai.VideoMetadata{FPS: genai.Ptr(0.5)}
parts := []*genai.Part{
videoPart,
genai.NewPartFromText("Generate a poster image that captures the key themes of this video."),
}
contents := []*genai.Content{
genai.NewContentFromParts(parts, genai.RoleUser),
}
result, err := client.Models.GenerateContent(
ctx,
"gemini-3.1-flash-image",
contents,
&genai.GenerateContentConfig{
ResponseModalities: []string{"TEXT", "IMAGE"},
},
)
if err != nil {
log.Fatal(err)
}
for _, part := range result.Candidates[0].Content.Parts {
if part.InlineData != nil {
imageBytes := part.InlineData.Data
_ = os.WriteFile("video_poster.png", imageBytes, 0644)
log.Println("Image saved as video_poster.png")
}
}
}
Java
import com.google.genai.Client;
import com.google.genai.types.Content;
import com.google.genai.types.FileData;
import com.google.genai.types.GenerateContentConfig;
import com.google.genai.types.GenerateContentResponse;
import com.google.genai.types.Part;
import com.google.genai.types.VideoMetadata;
import com.google.common.collect.ImmutableList;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Paths;
public class VideoToImage {
public static void main(String[] args) throws IOException {
try (Client client = new Client()) {
Part videoPart = Part.builder()
.fileData(FileData.builder()
.fileUri("https://www.youtube.com/watch?v=UTdfxFyOQTI")
.build())
.videoMetadata(VideoMetadata.builder()
.fps(0.5)
.build())
.build();
Part textPart = Part.builder()
.text("Generate a poster image that captures the key themes of this video.")
.build();
GenerateContentConfig config = GenerateContentConfig.builder()
.responseModalities("TEXT", "IMAGE")
.build();
GenerateContentResponse response = client.models.generateContent(
"gemini-3.1-flash-image",
Content.builder()
.role("user")
.parts(ImmutableList.of(videoPart, textPart))
.build(),
config);
for (Part part : response.parts()) {
if (part.inlineData().isPresent()) {
var blob = part.inlineData().get();
if (blob.data().isPresent()) {
Files.write(Paths.get("video_poster.png"), blob.data().get());
System.out.println("Image saved as video_poster.png");
}
}
}
}
}
}
C#
using Google.GenAI;
using Google.GenAI.Types;
using System;
using System.Collections.Generic;
using System.IO;
using System.Threading.Tasks;
public class VideoToImage {
public static async Task Main(string[] args) {
var client = new Client();
var response = await client.Models.GenerateContentAsync(
model: "gemini-3.1-flash-image",
contents: new List<Part>
{
new Part
{
FileData = new FileData { FileUri = "https://www.youtube.com/watch?v=UTdfxFyOQTI" },
VideoMetadata = new VideoMetadata { Fps = 0.5 }
},
new Part { Text = "Generate a poster image that captures the key themes of this video." }
},
config: new GenerateContentConfig
{
ResponseModalities = new List<string> { "TEXT", "IMAGE" }
}
);
foreach (var candidate in response.Candidates) {
foreach (var part in candidate.Content.Parts) {
if (part.InlineData != null) {
var imageBytes = Convert.FromBase64String(part.InlineData.Data);
await File.WriteAllBytesAsync("video_poster.png", imageBytes);
Console.WriteLine("Image saved as video_poster.png");
}
}
}
}
}
REST
curl -s -X POST \
"https://generativelanguage.googleapis.com/v1/models/gemini-3.1-flash-image:generateContent" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"contents": [{
"parts": [
{
"file_data": {
"file_uri": "https://www.youtube.com/watch?v=UTdfxFyOQTI"
},
"video_metadata": {
"fps": 0.5
}
},
{"text": "Generate a poster image that captures the key themes of this video."}
]
}],
"generationConfig": {
"responseModalities": ["TEXT", "IMAGE"]
}
}'
生成最高 4K 解析度的圖片
Gemini 3 圖像模型預設會生成 1K 圖像，但也能輸出 2K、4K 和 512 (0.5K) 圖像 (僅限 Gemini 3.1 Flash Image)。如要產生高解析度素材資源，請在 generation_config 中指定 image_size。
必須使用大寫的「K」(例如 1K、2K、4K)。512 值不會使用「K」後置字元。系統會拒絕小寫參數 (例如 1k)。
Python
from google import genai
from google.genai import types
prompt = "Da Vinci style anatomical sketch of a dissected Monarch butterfly. Detailed drawings of the head, wings, and legs on textured parchment with notes in English."
aspect_ratio = "1:1" # "1:1","1:4","1:8","2:3","3:2","3:4","4:1","4:3","4:5","5:4","8:1","9:16","16:9","21:9"
resolution = "1K" # "512", "1K", "2K", "4K"
client = genai.Client()
response = client.models.generate_content(
model="gemini-3.1-flash-image",
contents=prompt,
config=types.GenerateContentConfig(
response_modalities=['TEXT', 'IMAGE'],
response_format={"image": {aspect_ratio: aspect_ratio, image_size: resolution}},
)
)
for part in response.parts:
if part.text is not None:
print(part.text)
elif image:= part.as_image():
image.save("butterfly.png")
JavaScript
import { GoogleGenAI } from "@google/genai";
import * as fs from "node:fs";
async function main() {
const ai = new GoogleGenAI({});
const prompt =
'Da Vinci style anatomical sketch of a dissected Monarch butterfly. Detailed drawings of the head, wings, and legs on textured parchment with notes in English.';
const aspectRatio = '1:1';
const resolution = '1K';
const response = await ai.models.generateContent({
model: 'gemini-3.1-flash-image',
contents: prompt,
config: {
responseModalities: ['TEXT', 'IMAGE'],
responseFormat: {
image: {
aspectRatio: aspectRatio,
imageSize: resolution,
}
},
},
});
for (const part of response.candidates[0].content.parts) {
if (part.text) {
console.log(part.text);
} else if (part.inlineData) {
const imageData = part.inlineData.data;
const buffer = Buffer.from(imageData, "base64");
fs.writeFileSync("image.png", buffer);
console.log("Image saved as image.png");
}
}
}
main();
Go
package main
import (
"context"
"fmt"
"log"
"os"
"google.golang.org/genai"
)
func main() {
ctx := context.Background()
client, err := genai.NewClient(ctx, nil)
if err != nil {
log.Fatal(err)
}
defer client.Close()
model := client.GenerativeModel("gemini-3.1-flash-image")
model.GenerationConfig = &pb.GenerationConfig{
ResponseModalities: []pb.ResponseModality{genai.Text, genai.Image},
ImageConfig: &pb.ImageConfig{
AspectRatio: "1:1",
ImageSize: "1K",
},
}
prompt := "Da Vinci style anatomical sketch of a dissected Monarch butterfly. Detailed drawings of the head, wings, and legs on textured parchment with notes in English."
resp, err := model.GenerateContent(ctx, genai.Text(prompt))
if err != nil {
log.Fatal(err)
}
for _, part := range resp.Candidates[0].Content.Parts {
if txt, ok := part.(genai.Text); ok {
fmt.Printf("%s", string(txt))
} else if img, ok := part.(genai.ImageData); ok {
err := os.WriteFile("butterfly.png", img.Data, 0644)
if err != nil {
log.Fatal(err)
}
}
}
}
Java
import com.google.genai.Client;
import com.google.genai.types.GenerateContentConfig;
import com.google.genai.types.GenerateContentResponse;
import com.google.genai.types.GoogleSearch;
import com.google.genai.types.ImageConfig;
import com.google.genai.types.Part;
import com.google.genai.types.Tool;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Paths;
public class HiRes {
public static void main(String[] args) throws IOException {
try (Client client = new Client()) {
GenerateContentConfig config = GenerateContentConfig.builder()
.responseModalities("TEXT", "IMAGE")
.imageConfig(ImageConfig.builder()
.aspectRatio("16:9")
.imageSize("4K")
.build())
.build();
GenerateContentResponse response = client.models.generateContent(
"gemini-3.1-flash-image", """
Da Vinci style anatomical sketch of a dissected Monarch butterfly.
Detailed drawings of the head, wings, and legs on textured
parchment with notes in English.
""",
config);
for (Part part : response.parts()) {
if (part.text().isPresent()) {
System.out.println(part.text().get());
} else if (part.inlineData().isPresent()) {
var blob = part.inlineData().get();
if (blob.data().isPresent()) {
Files.write(Paths.get("butterfly.png"), blob.data().get());
}
}
}
}
}
}
C#
using Google.GenAI;
using System;
using System.Collections.Generic;
using System.IO;
using System.Threading.Tasks;
public class HiRes {
public static async Task Main(string[] args) {
var client = new Client();
var response = await client.Models.GenerateContentAsync(
model: "gemini-3.1-flash-image",
contents: new List<Part>
{
new Part { Text = "Da Vinci style anatomical sketch of a dissected Monarch butterfly. Detailed drawings of the head, wings, and legs on textured parchment with notes in English." }
},
config: new GenerateContentConfig
{
ResponseModalities = new List<string> { "TEXT", "IMAGE" },
ImageConfig = new ImageConfig
{
AspectRatio = "1:1",
ImageSize = "1K"
}
}
);
foreach (var candidate in response.Candidates) {
foreach (var part in candidate.Content.Parts) {
if (part.Text != null) {
Console.WriteLine(part.Text);
} else if (part.InlineData != null) {
var imageBytes = Convert.FromBase64String(part.InlineData.Data);
await File.WriteAllBytesAsync("butterfly.png", imageBytes);
Console.WriteLine("Image saved as butterfly.png");
}
}
}
}
}
REST
curl -s -X POST \
"https://generativelanguage.googleapis.com/v1/models/gemini-3.1-flash-image:generateContent" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"contents": [{"parts": [{"text": "Da Vinci style anatomical sketch of a dissected Monarch butterfly. Detailed drawings of the head, wings, and legs on textured parchment with notes in English."}]}],
"tools": [{"google_search": {}}],
"generationConfig": {
"responseModalities": ["TEXT", "IMAGE"],
"responseFormat": {
"image": {"aspectRatio": "1:1", "imageSize": "1K"}
}
}
}'
以下是根據這項提示生成的圖片範例：
思考過程
Gemini 3 圖像模型是思考型模型，會使用推論程序 (「思考」) 處理複雜的提示。這項功能預設為啟用，且無法在 API 中停用。如要進一步瞭解思考過程，請參閱 Gemini 思考指南。
模型最多會生成兩張暫時圖片，測試構圖和邏輯。「思考」中的最後一張圖片也是最終轉譯圖片。
你可以查看生成最終圖像的過程。
Python
for part in response.parts:
if part.thought:
if part.text:
print(part.text)
elif image:= part.as_image():
image.show()
JavaScript
for (const part of response.candidates[0].content.parts) {
if (part.thought) {
if (part.text) {
console.log(part.text);
} else if (part.inlineData) {
const imageData = part.inlineData.data;
const buffer = Buffer.from(imageData, 'base64');
fs.writeFileSync('image.png', buffer);
console.log('Image saved as image.png');
}
}
}
Java
for (Part part : response.parts()) {
if (part.thought().orElse(false)) {
if (part.text().isPresent()) {
System.out.println(part.text().get());
} else if (part.inlineData().isPresent()) {
var blob = part.inlineData().get();
if (blob.data().isPresent()) {
Files.write(Paths.get("image.png"), blob.data().get());
System.out.println("Image saved as image.png");
}
}
}
}
C#
foreach (var candidate in response.Candidates) {
foreach (var part in candidate.Content.Parts) {
if (part.Thought) {
if (part.Text != null) {
Console.WriteLine(part.Text);
} else if (part.InlineData != null) {
var imageBytes = Convert.FromBase64String(part.InlineData.Data);
await File.WriteAllBytesAsync("image.png", imageBytes);
Console.WriteLine("Image saved as image.png");
}
}
}
}
控管思考程度
使用 Gemini 3.1 Flash Image 時，你可以控制模型思考的程度，在品質和延遲之間取得平衡。預設 thinkingLevel 為 minimal，支援的層級為 minimal 和 high。將 thinkingLevel 設為 minimal 可提供延遲時間最短的回應。請注意，
「最少思考」並不代表模型完全不會思考。
您可以新增 includeThoughts 布林值，決定是否要在回覆中傳回模型生成的想法，或將其隱藏。
Python
from google import genai
response = client.models.generate_content(
model="gemini-3.1-flash-image",
contents="A futuristic city built inside a giant glass bottle floating in space",
config=types.GenerateContentConfig(
response_modalities=["IMAGE"],
thinking_config=types.ThinkingConfig(
thinking_level="High",
include_thoughts=True
),
)
)
for part in response.parts:
if part.thought: # Skip outputting thoughts
continue
if part.text:
display(Markdown(part.text))
elif image:= part.as_image():
image.show()
JavaScript
import { GoogleGenAI } from "@google/genai";
import * as fs from "node:fs";
async function main() {
const ai = new GoogleGenAI({});
const response = await ai.models.generateContent({
model: "gemini-3.1-flash-image",
contents: "A futuristic city built inside a giant glass bottle floating in space",
config: {
responseModalities: ["IMAGE"],
thinkingConfig: {
thinkingLevel: "High",
includeThoughts: true
},
},
});
for (const part of response.candidates[0].content.parts) {
if (part.thought) { // Skip outputting thoughts
continue;
}
if (part.text) {
console.log(part.text);
} else if (part.inlineData) {
const imageData = part.inlineData.data;
const buffer = Buffer.from(imageData, "base64");
fs.writeFileSync("image.png", buffer);
console.log("Image saved as image.png");
}
}
}
main();
Go
package main
import (
"context"
"fmt"
"log"
"os"
"google.golang.org/genai"
pb "google.golang.org/genai/schema"
)
func main() {
ctx := context.Background()
client, err := genai.NewClient(ctx, nil)
if err != nil {
log.Fatal(err)
}
defer client.Close()
model := client.GenerativeModel("gemini-3.1-flash-image")
model.GenerationConfig = &pb.GenerationConfig{
ResponseModalities: []pb.ResponseModality{genai.Image},
ThinkingConfig: &pb.ThinkingConfig{
ThinkingLevel: "High",
IncludeThoughts: true,
},
}
prompt := "A futuristic city built inside a giant glass bottle floating in space"
resp, err := model.GenerateContent(ctx, genai.Text(prompt))
if err != nil {
log.Fatal(err)
}
for _, part := range resp.Candidates[0].Content.Parts {
if part.Thought { // Skip outputting thoughts
continue
}
if txt, ok := part.(genai.Text); ok {
fmt.Printf("%s", string(txt))
} else if img, ok := part.(genai.ImageData); ok {
err := os.WriteFile("image.png", img.Data, 0644)
if err != nil {
log.Fatal(err)
}
}
}
}
Java
import com.google.genai.Client;
import com.google.genai.types.GenerateContentConfig;
import com.google.genai.types.GenerateContentResponse;
import com.google.genai.types.Part;
import com.google.genai.types.ThinkingConfig;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Paths;
public class ThinkingLevels {
public static void main(String[] args) throws IOException {
try (Client client = new Client()) {
GenerateContentConfig config = GenerateContentConfig.builder()
.responseModalities("IMAGE")
.thinkingConfig(ThinkingConfig.builder()
.thinkingLevel("High")
.includeThoughts(true)
.build())
.build();
GenerateContentResponse response = client.models.generateContent(
"gemini-3.1-flash-image",
"A futuristic city built inside a giant glass bottle floating in space",
config);
for (Part part : response.parts()) {
if (part.thought().orElse(false)) {
// Skip outputting thoughts
continue;
}
if (part.text().isPresent()) {
System.out.println(part.text().get());
} else if (part.inlineData().isPresent()) {
var blob = part.inlineData().get();
if (blob.data().isPresent()) {
Files.write(Paths.get("image.png"), blob.data().get());
System.out.println("Image saved as image.png");
}
}
}
}
}
}
C#
using Google.GenAI;
using System;
using System.Collections.Generic;
using System.IO;
using System.Threading.Tasks;
public class ThinkingLevels {
public static async Task Main(string[] args) {
var client = new Client();
var response = await client.Models.GenerateContentAsync(
model: "gemini-3.1-flash-image",
contents: new List<Part>
{
new Part { Text = "A futuristic city built inside a giant glass bottle floating in space" }
},
config: new GenerateContentConfig
{
ResponseModalities = new List<string> { "IMAGE" },
ThinkingConfig = new ThinkingConfig
{
ThinkingLevel = "High",
IncludeThoughts = true
}
}
);
foreach (var candidate in response.Candidates) {
foreach (var part in candidate.Content.Parts) {
if (part.Thought) {
// Skip outputting thoughts
continue;
}
if (part.Text != null) {
Console.WriteLine(part.Text);
} else if (part.InlineData != null) {
var imageBytes = Convert.FromBase64String(part.InlineData.Data);
await File.WriteAllBytesAsync("image.png", imageBytes);
Console.WriteLine("Image saved as image.png");
}
}
}
}
}
REST
curl -s -X POST \
"https://generativelanguage.googleapis.com/v1/models/gemini-3.1-flash-image:generateContent" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"contents": [{"parts": [{"text": "A futuristic city built inside a giant glass bottle floating in space"}]}],
"generationConfig": {
"responseModalities": ["IMAGE"],
"thinkingConfig": {
"thinkingLevel": "High",
"includeThoughts": true
}
}
}'
請注意，無論 includeThoughts 設為 true 或 false，系統都會收取思考權杖費用，因為無論您是否查看思考過程，系統預設都會執行思考程序。
想法簽名
思想簽章是模型內部思考過程的加密表示法，用於在多輪互動中保留推理脈絡。所有回應都包含 thought_signature 欄位。一般來說，如果在模型回覆中收到想法簽章，您應該在下一個回合傳送對話記錄時，完全按照收到的內容傳回該簽章。如果無法傳送思維簽章，可能會導致回應失敗。如要進一步瞭解簽章，請參閱想法簽章說明文件。
思想簽章的運作方式如下：
- 回覆中所有含有圖片的
inline_data部分mimetype都應有簽名。 - 如果想法之後緊接著有文字部分 (在任何圖片之前)，則第一個文字部分也應包含簽名。
- 如果
inline_data含有圖片的mimetype部分是想法的一部分，就不會有簽名。
以下程式碼範例顯示包含想法簽章的位置：
[
{
"inline_data": {
"data": "<base64_image_data_0>",
"mime_type": "image/png"
},
"thought": true // Thoughts don't have signatures
},
{
"inline_data": {
"data": "<base64_image_data_1>",
"mime_type": "image/png"
},
"thought": true // Thoughts don't have signatures
},
{
"inline_data": {
"data": "<base64_image_data_2>",
"mime_type": "image/png"
},
"thought": true // Thoughts don't have signatures
},
{
"text": "Here is a step-by-step guide to baking macarons, presented in three separate images.\n\n### Step 1: Piping the Batter\n\nThe first step after making your macaron batter is to pipe it onto a baking sheet. This requires a steady hand to create uniform circles.\n\n",
"thought_signature": "<Signature_A>" // The first non-thought part always has a signature
},
{
"inline_data": {
"data": "<base64_image_data_3>",
"mime_type": "image/png"
},
"thought_signature": "<Signature_B>" // All image parts have a signatures
},
{
"text": "\n\n### Step 2: Baking and Developing Feet\n\nOnce piped, the macarons are baked in the oven. A key sign of a successful bake is the development of \"feet\"—the ruffled edge at the base of each macaron shell.\n\n"
// Follow-up text parts don't have signatures
},
{
"inline_data": {
"data": "<base64_image_data_4>",
"mime_type": "image/png"
},
"thought_signature": "<Signature_C>" // All image parts have a signatures
},
{
"text": "\n\n### Step 3: Assembling the Macaron\n\nThe final step is to pair the cooled macaron shells by size and sandwich them together with your desired filling, creating the classic macaron dessert.\n\n"
},
{
"inline_data": {
"data": "<base64_image_data_5>",
"mime_type": "image/png"
},
"thought_signature": "<Signature_D>" // All image parts have a signatures
}
]
其他圖像生成模式
Gemini 支援其他圖像互動模式，包括：
- 文字生成圖片和文字 (交錯)：輸出圖片和相關文字。
- 提示範例：「Generate an illustrated recipe for a paella.」(生成西班牙海鮮飯的插圖食譜)。
- 圖片和文字轉圖片和文字 (交錯)：使用輸入的圖片和文字，建立新的相關圖片和文字。
- 提示範例：(附上附家具的房間圖片)「我的空間還適合什麼顏色的沙發？可以更新圖片嗎？」
批次生成圖片
如需產生大量圖片，可以使用批次 API。你可換取更高的速率限制，但須等待最多 24 小時。
請參閱批次 API 圖片生成說明文件和食譜，瞭解批次 API 圖片範例和程式碼。
提示撰寫指南和策略
如要精通圖像生成，首先要掌握一項基本原則：
描述場景，不要只列出關鍵字。 這項模型的核心優勢在於深入理解語言，與不相關的字詞清單相比，敘事性描述段落幾乎都能產生更優質、更連貫的圖片。
生成圖片的提示
下列策略可協助您撰寫有效提示，生成所需圖片。
攝影
如要生成真實感十足的圖片，請使用攝影術語。提及相機角度、鏡頭類型、光線和細節，引導模型生成寫實的結果。
| 提示 | 生成內容 |
|---|---|
| 相片：一位年長的日本陶藝家，臉上刻滿深刻的皺紋，面帶溫暖的微笑，他正在仔細檢查剛上釉的茶碗。場景是陽光充足的鄉村風工作室。窗外灑入柔和的黃金時段光線，照亮整個場景，凸顯黏土的細緻紋理。使用 85mm 人像鏡頭拍攝，呈現柔和模糊的背景 (散景)。整體氛圍寧靜且充滿大師風範。直向。 |
|
風格插圖和貼圖
如要製作貼紙、圖示或素材資源，請明確指定樣式並要求白色背景。
| 提示 | 生成內容 |
|---|---|
| 可愛風格的貼紙：一隻戴著小竹帽的開心小熊貓。正在啃食綠色竹葉。設計特色包括大膽簡潔的輪廓、簡單的賽璐珞陰影，以及鮮豔的色調。背景必須為白色。 |
|
圖片中的文字準確
Gemini 擅長算繪文字，清楚說明文字、字型樣式 (描述性) 和整體設計。使用 Gemini 3 Pro Image 製作專業素材。
| 提示 | 生成內容 |
|---|---|
| 為名為「The Daily Grind」的咖啡廳設計現代極簡風格的標誌。文字應使用簡潔的粗體 Sans Serif 字型。色彩配置為黑白。將標誌放在圓圈中。善用咖啡豆。 |
|
產品模型和商業攝影
非常適合為電子商務、廣告或品牌宣傳製作乾淨俐落的專業產品照片。
| 提示 | 生成內容 |
|---|---|
| 高解析度攝影棚產品照：霧面黑色極簡陶瓷咖啡杯，放在拋光混凝土表面上。燈光是三點柔光箱設定，旨在營造柔和的漫射高光，並消除強烈的陰影。攝影機角度略為抬高，以 45 度拍攝，展現俐落線條。極度逼真，並清楚呈現咖啡冒出的蒸氣。正方形圖片。 |
|
極簡風和負空間設計
非常適合用於建立網站、簡報或行銷素材的背景，並在上面疊加文字。
| 提示 | 生成內容 |
|---|---|
| 極簡構圖，畫面右下角有一片精緻的紅楓葉。背景是廣闊的米白色空白畫布，為文字創造出大量的負空間。左上角柔和的漫射光源。正方形圖片。 |
|
連續圖像 (漫畫格 / 分鏡腳本)
以角色一致性和場景描述為基礎，為視覺故事創作建立面板。如要確保文字準確度和敘事能力，建議使用 Gemini 3.1 Pro 和 Gemini 3.1 Flash Image 撰寫提示。
| 提示 | 生成內容 |
|---|---|
|
輸入圖片： 
提示：製作 3 格漫畫，採用高對比黑白墨水，呈現粗獷的黑色電影藝術風格。將角色置於幽默的場景中。 |
|
以 Google 搜尋建立基準
使用 Google 搜尋，根據近期或即時資訊生成圖片。 這項功能適用於新聞、天氣和其他時效性主題。
| 提示 | 生成內容 |
|---|---|
| 製作昨晚阿森納在歐洲冠軍聯賽的賽事圖表，風格簡約但時尚 |
|
編輯圖片的提示
這些範例說明如何提供圖片和文字提示，以進行編輯、合成和風格轉移。
新增及移除元素
提供圖片並說明想做的變更。模型會與原始圖片的風格、光線和透視效果相符。
| 提示 | 生成內容 |
|---|---|
|
輸入圖片： 
提示：請使用我提供的貓咪圖片，在貓咪頭上加上一頂小小的針織魔法師帽。讓寵物看起來舒適自在，並與相片的柔和光線相符。 |
|
局部重繪 (語意遮罩)
以對話方式定義「遮罩」，編輯圖像的特定部分，其餘部分則保持不變。
| 提示 | 生成內容 |
|---|---|
|
輸入圖片： 
提示：使用提供的客廳圖片，只將藍色沙發換成復古的棕色皮革切斯特菲爾德沙發。保持房間其他部分不變，包括沙發上的枕頭和燈光。 |
|
風格轉換
提供圖片，要求模型以不同藝術風格重新創作內容。
| 提示 | 生成內容 |
|---|---|
|
輸入圖片： 
提示：將提供的現代城市街道夜景相片，轉換成梵谷《星夜》的藝術風格。保留建築物和車輛的原始構圖，但以旋轉的厚塗筆觸和深藍色與亮黃色的戲劇性調色盤，呈現所有元素。 |
|
進階合成：合併多張圖片
提供多張圖片做為情境，建立新的複合場景。這項功能非常適合製作產品模型或創意拼貼。
| 提示 | 生成內容 |
|---|---|
|
輸入圖片： 
提示：製作專業的電子商務時尚相片。將第一張圖片中的藍色花卉洋裝，套用在第二張圖片中的女性身上。生成一張寫實的全身照，圖中女子穿著這件洋裝，並根據戶外環境調整光線和陰影。 |
|
保留高保真細節
為確保編輯時保留重要細節 (例如臉部或標誌)，請在編輯要求中詳細說明這些細節。
| 提示 | 生成內容 |
|---|---|
|
輸入圖片： 
提示：以第一張圖片為範本，生成一名棕髮、藍眼、表情中性的女子。將第二張圖片中的標誌加到她穿的黑色 T 恤上。請確保女性的臉部和特徵完全不變。標誌應看起來像是自然印在布料上，並隨著襯衫的摺疊而彎曲。 |
|
讓某個事物活起來
上傳草圖或手繪圖，要求模型將其精修成完成圖。
| 提示 | 生成內容 |
|---|---|
|
輸入圖片： 
提示：將這張未來汽車的鉛筆草圖，變成展示間中精緻的最終概念車照片。保留草圖中的俐落線條和低調外觀，但加入金屬藍色塗裝和霓虹燈環繞照明。 |
|
角色一致性：360 度全景
您可以透過反覆提示不同角度，生成角色的 360 度視角。為獲得最佳效果，請在後續的提示中加入先前生成的圖片，以維持一致性。如要生成複雜姿勢，請附上所需姿勢的參考圖像。
| 提示 | 生成內容 |
|---|---|
|
輸入圖片：
提示：這名男子在白色背景前拍攝的攝影棚肖像照，側臉朝右 |
|
最佳做法
如要讓成果更上一層樓，請在工作流程中加入這些專業策略。
- 具體說明：提供的細節越多，你對生成結果的掌控權就越大。請描述「奇幻盔甲」，例如「裝飾華麗的精靈板甲，刻有銀葉圖案，高領，肩甲形狀像獵鷹翅膀」。
- 提供背景資訊和意圖：說明圖片的用途。模型對背景資訊的理解程度會影響最終輸出內容。舉例來說，「為高檔極簡護膚品牌設計標誌」比「設計標誌」更能產生理想結果。
- 反覆測試及修正：請勿期待第一次就能生成完美圖片。運用模型的對話功能進行小幅變更。接著輸入「很棒，但可以讓光線暖一點嗎？」或「維持所有設定，但將角色的表情改為更嚴肅。」等提示。
- 使用逐步指示：如果場景複雜且包含許多元素，請將提示分成多個步驟。「首先，請在黎明時分，建立一片寧靜、霧氣瀰漫的森林背景。接著，在前景中加入長滿青苔的古老石祭壇。 最後，將一把發光的劍放在祭壇上。」
- 使用「語意負面提示」：不要說「沒有車輛」，而是正面描述想要的場景：「空蕩蕩的荒涼街道，沒有任何交通跡象」。
- 控制相機：使用攝影和電影語言控制構圖。例如
wide-angle shot、macro shot、low-angle perspective。
限制
- 如要獲得最佳效能，請使用下列語言：英文、ar-EG、de-DE、es-MX、fr-FR、hi-IN、id-ID、it-IT、ja-JP、ko-KR、pt-BR、ru-RU、ua-UA、vi-VN、zh-CN。
- 圖像生成功能不支援音訊輸入內容。只有 Gemini 3.1 Flash Image 支援影片輸入。
- 模型不一定會輸出使用者明確要求的確切圖片數量。
gemini-2.5-flash-image最多可輸入 3 張圖片，gemini-3-pro-image則支援 5 張高保真度圖片，最多可輸入 14 張圖片。gemini-3.1-flash-image支援最多 4 個字元的相似度，以及單一工作流程中最多 10 個物件的保真度。- 為圖片生成文字時，建議先生成文字，然後要求提供含有該文字的圖片，這樣 Gemini 的效果最好。
gemini-3.1-flash-image目前，以 Google 搜尋強化事實基礎時，不支援使用網路搜尋中的人物真實圖像。- 所有生成的圖片都會加上 SynthID 浮水印。
選用設定
您可以在 generate_content 呼叫的 config 欄位中，視需要設定模型輸出內容的回應模式和長寬比。
輸出類型
模型預設會傳回文字和圖片回應 (即 response_modalities=['Text', 'Image'])。您可以透過 response_modalities=['Image'] 設定回應，只傳回圖片，不傳回文字。
Python
response = client.models.generate_content(
model="gemini-3.1-flash-image",
contents=[prompt],
config=types.GenerateContentConfig(
response_modalities=['Image']
)
)
JavaScript
const response = await ai.models.generateContent({
model: "gemini-3.1-flash-image",
contents: prompt,
config: {
responseModalities: ['Image']
}
});
Go
result, _ := client.Models.GenerateContent(
ctx,
"gemini-3.1-flash-image",
genai.Text("Create a picture of a nano banana dish in a " +
" fancy restaurant with a Gemini theme"),
&genai.GenerateContentConfig{
ResponseModalities: "Image",
},
)
Java
response = client.models.generateContent(
"gemini-3.1-flash-image",
prompt,
GenerateContentConfig.builder()
.responseModalities("IMAGE")
.build());
C#
var response = await client.Models.GenerateContentAsync(
model: "gemini-3.1-flash-image",
contents: new List<Part> { new Part { Text = prompt } },
config: new GenerateContentConfig
{
ResponseModalities = new List<string> { "IMAGE" }
}
);
REST
curl -s -X POST \
"https://generativelanguage.googleapis.com/v1/models/gemini-3.1-flash-image:generateContent" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"contents": [{
"parts": [
{"text": "Create a picture of a nano banana dish in a fancy restaurant with a Gemini theme"}
]
}],
"generationConfig": {
"responseModalities": ["Image"]
}
}'
顯示比例和圖片大小
模型預設會將輸出圖片大小設為與輸入圖片相同，否則會生成 1:1 的正方形。
您可以使用回應要求中 response_format 下方的 aspect_ratio 欄位，控制輸出圖片的顯示比例，如下所示：
Python
# For gemini-2.5-flash-image
response = client.models.generate_content(
model="gemini-2.5-flash-image",
contents=[prompt],
config=types.GenerateContentConfig(
response_format={"image": {aspect_ratio: "16:9",}}
)
)
# For gemini-3.1-flash-image and gemini-3-pro-image
response = client.models.generate_content(
model="gemini-3.1-flash-image",
contents=[prompt],
config=types.GenerateContentConfig(
response_format={"image": {aspect_ratio: "16:9", image_size: "2K",}}
)
)
JavaScript
// For gemini-2.5-flash-image
const response = await ai.models.generateContent({
model: "gemini-2.5-flash-image",
contents: prompt,
config: {
responseFormat: {
image: {
aspectRatio: "16:9",
}
},
}
});
// For gemini-3.1-flash-image and gemini-3-pro-image
const response_gemini3 = await ai.models.generateContent({
model: "gemini-3.1-flash-image",
contents: prompt,
config: {
responseFormat: {
image: {
aspectRatio: "16:9",
imageSize: "2K",
}
},
}
});
Go
// For gemini-2.5-flash-image
result, _ := client.Models.GenerateContent(
ctx,
"gemini-2.5-flash-image",
genai.Text("Create a picture of a nano banana dish in a " +
" fancy restaurant with a Gemini theme"),
&genai.GenerateContentConfig{
ImageConfig: &genai.ImageConfig{
AspectRatio: "16:9",
},
}
)
// For gemini-3.1-flash-image and gemini-3-pro-image
result_gemini3, _ := client.Models.GenerateContent(
ctx,
"gemini-3.1-flash-image",
genai.Text("Create a picture of a nano banana dish in a " +
" fancy restaurant with a Gemini theme"),
&genai.GenerateContentConfig{
ImageConfig: &genai.ImageConfig{
AspectRatio: "16:9",
ImageSize: "2K",
},
}
)
Java
// For gemini-2.5-flash-image
response = client.models.generateContent(
"gemini-2.5-flash-image",
prompt,
GenerateContentConfig.builder()
.imageConfig(ImageConfig.builder()
.aspectRatio("16:9")
.build())
.build());
// For gemini-3.1-flash-image and gemini-3-pro-image
response_gemini3 = client.models.generateContent(
"gemini-3.1-flash-image",
prompt,
GenerateContentConfig.builder()
.imageConfig(ImageConfig.builder()
.aspectRatio("16:9")
.imageSize("2K")
.build())
.build());
C#
// For gemini-2.5-flash-image
var response = await client.Models.GenerateContentAsync(
model: "gemini-2.5-flash-image",
contents: new List<Part> { new Part { Text = prompt } },
config: new GenerateContentConfig
{
ImageConfig = new ImageConfig
{
AspectRatio = "16:9"
}
}
);
// For gemini-3.1-flash-image and gemini-3-pro-image
var response_gemini3 = await client.Models.GenerateContentAsync(
model: "gemini-3.1-flash-image",
contents: new List<Part> { new Part { Text = prompt } },
config: new GenerateContentConfig
{
ImageConfig = new ImageConfig
{
AspectRatio = "16:9",
ImageSize = "2K"
}
}
);
REST
# For gemini-2.5-flash-image
curl -s -X POST \
"https://generativelanguage.googleapis.com/v1/models/gemini-2.5-flash-image:generateContent" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H 'Content-Type: application/json' \
-d '{
"contents": [{
"parts": [
{"text": "Create a picture of a nano banana dish in a fancy restaurant with a Gemini theme"}
]
}],
"generationConfig": {
"responseFormat": {
"image": {
"aspectRatio": "16:9"
}
}
}
}'
# For gemini-3-pro-image
curl -s -X POST \
"https://generativelanguage.googleapis.com/v1/models/gemini-3.1-flash-image:generateContent" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H 'Content-Type: application/json' \
-d '{
"contents": [{
"parts": [
{"text": "Create a picture of a nano banana dish in a fancy restaurant with a Gemini theme"}
]
}],
"generationConfig": {
"responseFormat": {
"image": {
"aspectRatio": "16:9",
"imageSize": "2K"
}
}
}
}'
下表列出可用的不同比例和生成的圖片大小：
3.1 Flash Image
| 顯示比例 | 512 解析度 | 500 個權杖 | 1K 解析度 | 1,000 個權杖 | 2K 解析度 | 2,000 個權杖 | 4K 解析度 | 4,000 個權杖 |
|---|---|---|---|---|---|---|---|---|
| 1:1 | 512x512 | 747 | 1024x1024 | 1120 | 2048 x 2048 | 1680 | 4096x4096 | 2520 |
| 1:4 | 256x1024 | 747 | 512x2048 | 1120 | 1024x4096 | 1680 | 2048x8192 | 2520 |
| 1:8 | 192x1536 | 747 | 384x3072 | 1120 | 768x6144 | 1680 | 1536x12288 | 2520 |
| 2:3 | 424x632 | 747 | 848x1264 | 1120 | 1696x2528 | 1680 | 3392x5056 | 2520 |
| 3:2 | 632x424 | 747 | 1264x848 | 1120 | 2528x1696 | 1680 | 5056x3392 | 2520 |
| 3:4 | 448x600 | 747 | 896x1200 | 1120 | 1792x2400 | 1680 | 3584x4800 | 2520 |
| 4:1 | 1024x256 | 747 | 2048x512 | 1120 | 4096x1024 | 1680 | 8192x2048 | 2520 |
| 4:3 | 600x448 | 747 | 1200x896 | 1120 | 2400x1792 | 1680 | 4800x3584 | 2520 |
| 4:5 | 464x576 | 747 | 928x1152 | 1120 | 1856x2304 | 1680 | 3712x4608 | 2520 |
| 5:4 | 576x464 | 747 | 1152x928 | 1120 | 2304x1856 | 1680 | 4608x3712 | 2520 |
| 8:1 | 1536x192 | 747 | 3072x384 | 1120 | 6144x768 | 1680 | 12288x1536 | 2520 |
| 9:16 | 384x688 | 747 | 768x1376 | 1120 | 1536x2752 | 1680 | 3072x5504 | 2520 |
| 16:9 | 688x384 | 747 | 1376x768 | 1120 | 2752x1536 | 1680 | 5504x3072 | 2520 |
| 21:9 | 792x168 | 747 | 1584x672 | 1120 | 3168x1344 | 1680 | 6336x2688 | 2520 |
3.1 Pro Image
| 顯示比例 | 1K 解析度 | 1,000 個權杖 | 2K 解析度 | 2,000 個權杖 | 4K 解析度 | 4,000 個權杖 |
|---|---|---|---|---|---|---|
| 1:1 | 1024x1024 | 1120 | 2048 x 2048 | 1120 | 4096x4096 | 2000 |
| 2:3 | 848x1264 | 1120 | 1696x2528 | 1120 | 3392x5056 | 2000 |
| 3:2 | 1264x848 | 1120 | 2528x1696 | 1120 | 5056x3392 | 2000 |
| 3:4 | 896x1200 | 1120 | 1792x2400 | 1120 | 3584x4800 | 2000 |
| 4:3 | 1200x896 | 1120 | 2400x1792 | 1120 | 4800x3584 | 2000 |
| 4:5 | 928x1152 | 1120 | 1856x2304 | 1120 | 3712x4608 | 2000 |
| 5:4 | 1152x928 | 1120 | 2304x1856 | 1120 | 4608x3712 | 2000 |
| 9:16 | 768x1376 | 1120 | 1536x2752 | 1120 | 3072x5504 | 2000 |
| 16:9 | 1376x768 | 1120 | 2752x1536 | 1120 | 5504x3072 | 2000 |
| 21:9 | 1584x672 | 1120 | 3168x1344 | 1120 | 6336x2688 | 2000 |
Gemini 2.5 Flash Image
| 顯示比例 | 解析度 | 符記數 |
|---|---|---|
| 1:1 | 1024x1024 | 1290 |
| 2:3 | 832x1248 | 1290 |
| 3:2 | 1248x832 | 1290 |
| 3:4 | 864x1184 | 1290 |
| 4:3 | 1184x864 | 1290 |
| 4:5 | 896x1152 | 1290 |
| 5:4 | 1152x896 | 1290 |
| 9:16 | 768x1344 | 1290 |
| 16:9 | 1344x768 | 1290 |
| 21:9 | 1536x672 | 1290 |
多種模型供您選擇
選擇最適合特定用途的模型。
Gemini 3.1 Flash Image (Nano Banana 2) 是您圖像生成模型的首選，因為它在效能、智慧、成本和延遲之間取得最佳平衡。詳情請參閱模型定價和功能頁面。
Gemini 3 Pro Image (Nano Banana Pro) 專為專業資產製作和複雜指令而設計。這個模型會使用 Google 搜尋功能，以現實世界為基礎生成圖像，並預設在生成圖像前先「思考」，以改善構圖，且能生成最高 4K 解析度的圖像。詳情請參閱模型定價和功能頁面。
Gemini 2.5 Flash Image (Nano Banana) 的設計宗旨是速度和效率。這個模型經過最佳化，可處理大量低延遲工作，並生成 1024 像素的圖片。詳情請參閱模型定價和功能頁面。
Imagen 的適用時機
除了使用 Gemini 內建的圖像生成功能，你也可以透過 Gemini API 存取專門的圖像生成模型 Imagen。請務必在關閉日期前完成遷移。