Image generation

Image generation (이미지 생성)

이미지 생성 도구는 텍스트 프롬프트와, 선택적으로 이미지 입력을 사용해 이미지를 생성하게 해줘요. gpt-image-2.5-sunburst, gpt-image-2.5-flare, gpt-image-2, gpt-image-1.5, gpt-image-1, gpt-image-1-mini 같은 GPT Image 모델을 사용하며, 성능 향상을 위해 텍스트 입력을 자동으로 최적화해요.

image_generation 도구의 model을 정밀 편집에는 gpt-image-2.5-sunburst로, 빠르고 고품질의 이미지 생성에는 gpt-image-2.5-flare로 설정하세요. 최상위 Responses model 필드에는 지원되는 메인라인 모델(mainline model)을 사용하세요.

이미지 생성에 대해 더 알아보고 싶다면 전용 이미지 생성 가이드를 참고하세요.

출처: 문서

본문

사용법

요청에 image_generation 도구를 포함하면, 모델이 대화의 일부로 언제·어떻게 이미지를 생성할지 프롬프트와 제공된 이미지 입력을 바탕으로 결정할 수 있어요.

image_generation_call 도구 호출 결과에는 base64로 인코딩된 이미지가 포함돼요.

이미지 생성

import OpenAI from "openai";
const openai = new OpenAI();

const response = await openai.responses.create({
  model: "gpt-6-astra",
  input:
    "Generate an image of gray tabby cat hugging an otter with an orange scarf",
  tools: [{ type: "image_generation", model: "gpt-image-2.5-sunburst" }],
});

// Save the image to a file
const imageData = response.output
  .filter((output) => output.type === "image_generation_call")
  .map((output) => output.result);

if (imageData.length > 0) {
  const imageBase64 = imageData[0];
  const fs = await import("fs");
  fs.writeFileSync("otter.png", Buffer.from(imageBase64, "base64"));
}

from openai import OpenAI
import base64

client = OpenAI()

response = client.responses.create(
    model="gpt-6-astra",
    input="Generate an image of gray tabby cat hugging an otter with an orange scarf",
    tools=[{"type": "image_generation", "model": "gpt-image-2.5-sunburst"}],
)

# Save the image to a file
image_data = [
    output.result
    for output in response.output
    if output.type == "image_generation_call"
]

if image_data:
    image_base64 = image_data[0]
    with open("otter.png", "wb") as f:
        f.write(base64.b64decode(image_base64))

package main

import (
	"context"
	"encoding/base64"
	"os"

	"github.com/openai/openai-go/v3"
	"github.com/openai/openai-go/v3/responses"
)

func main() {
	client := openai.NewClient()
	response, err := client.Responses.New(context.Background(), responses.ResponseNewParams{
		Model: "gpt-6-astra",
		Input: responses.ResponseNewParamsInputUnion{
			OfString: openai.String("Generate an image of gray tabby cat hugging an otter with an orange scarf"),
		},
		Tools: []responses.ToolUnionParam{{OfImageGeneration: &responses.ToolImageGenerationParam{Model: "gpt-image-2.5-sunburst"}}},
	})
	if err != nil {
		panic(err)
	}
	saveFirstGeneratedImage(response, "otter.png")
}

func saveFirstGeneratedImage(response *responses.Response, filename string) {
	for _, output := range response.Output {
		if output.Type != "image_generation_call" {
			continue
		}
		image, err := base64.StdEncoding.DecodeString(output.AsImageGenerationCall().Result)
		if err != nil {
			panic(err)
		}
		if err := os.WriteFile(filename, image, 0o600); err != nil {
			panic(err)
		}
		return
	}
	panic("response did not include an image generation call")
}

import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.models.responses.ResponseCreateParams;
import com.openai.models.responses.Tool;
import java.nio.file.Files;
import java.nio.file.Path;
import java.util.Base64;

ResponseCreateParams params =
    ResponseCreateParams.builder()
        .model("gpt-6-astra")
        .input("Generate an image of a gray tabby cat hugging an otter with an orange scarf.")
        .addTool(Tool.ImageGeneration.builder().build())
        .build();

var image =
    client.responses().create(params).output().stream()
        .flatMap(item -> item.imageGenerationCall().stream())
        .findFirst()
        .orElseThrow(() -> new IllegalStateException("No image generation call returned"));
String encoded =
    image.result().orElseThrow(() -> new IllegalStateException("No image returned"));
Files.write(Path.of("otter.png"), Base64.getDecoder().decode(encoded));

using OpenAI.Responses;
#pragma warning disable OPENAI001

string key = Environment.GetEnvironmentVariable("OPENAI_API_KEY")!;
ResponsesClient client = new(key);

CreateResponseOptions options = new() { Model = "gpt-6-astra" };
options.InputItems.Add(
    ResponseItem.CreateUserMessageItem(
        "Generate an image of a gray tabby cat hugging an otter with an orange scarf."
    )
);
options.Tools.Add(ResponseTool.CreateImageGenerationTool(model: "gpt-image-2.5-sunburst"));

ResponseResult response = await client.CreateResponseAsync(options);
ImageGenerationCallResponseItem image = response
    .OutputItems.OfType<ImageGenerationCallResponseItem>()
    .FirstOrDefault()
    ?? throw new InvalidOperationException("No generated image was returned.");
await File.WriteAllBytesAsync("otter.png", image.ImageResultBytes.ToArray());

require "base64"
require "openai"

client = OpenAI::Client.new
response = client.responses.create(
  model: "gpt-6-astra",
  input: "Generate an image of a gray tabby cat hugging an otter with an orange scarf.",
  tools: [
    {
      type: :image_generation,
      model: "gpt-image-2.5-sunburst"
    }
  ]
)

image_call = response.output.find do |item|
  item.is_a?(OpenAI::Models::Responses::ResponseOutputItem::ImageGenerationCall)
end
unless image_call.is_a?(OpenAI::Models::Responses::ResponseOutputItem::ImageGenerationCall)
  raise "No image generation call returned"
end

encoded_image = image_call.result or raise "No image returned"
File.binwrite("otter.png", Base64.strict_decode64(encoded_image))

입력 이미지는 파일 ID나 base64 데이터로 제공할 수 있어요.

이미지 생성 도구 호출을 강제하려면 tool_choice 파라미터를 {"type": "image_generation"}으로 설정하면 돼요.

도구 옵션

이미지 생성 도구에 다음 출력 옵션을 파라미터로 설정할 수 있어요.

• Size: 이미지 크기 (예: 1024 × 1024 또는 1024 × 1536)
• Quality: 렌더링 품질 (예: low, medium, high)
• Format: 파일 출력 형식
• Compression: JPEG·WebP 형식의 압축 수준 (0-100%)
• Background: 투명(transparent), 불투명(opaque), 자동(automatic)
• Action: 요청이 이미지를 자동 선택, 생성, 편집 중 어떤 것을 할지

size, quality, background는 auto 옵션을 지원하며, 이 경우 모델이 프롬프트에 따라 최적의 옵션을 자동으로 선택해요.

gpt-image-2.5-sunburst와 gpt-image-2.5-flare의 경우 quality는 xhigh와 max도 받아요. 이런 값은 이전 GPT Image 모델에서는 지원되지 않아요. 기본 품질은 여전히 auto예요.

gpt-image-2는 해상도 제약 조건을 충족하는 유연한 size 값을 지원해요. 투명 배경은 프리뷰에서 사용할 수 있으며, 요청하려면 background: "transparent"를 설정하면 돼요. png(기본값) 또는 webp를 사용하고, 투명 배경에서는 jpeg가 지원되지 않아요.

사용 가능한 옵션에 대한 자세한 내용은 이미지 생성 가이드를 참고하세요.

Responses API 이미지 생성 도구를 쓰면 지원되는 GPT Image 모델이 새 이미지를 생성할지, 대화에 이미 있는 이미지를 편집할지 선택할 수 있어요. 선택적 action 파라미터가 이 동작을 제어하는데, action을 auto로 두면 모델이 생성·편집 중 하나를 고르고, generate 또는 edit로 설정하면 해당 동작을 강제해요. 지정하지 않으면 기본값은 auto예요.

수정된 프롬프트 (Revised prompt)

이미지 생성 도구를 사용할 때 메인라인 모델(예: gpt-5.5)이 성능 향상을 위해 프롬프트를 자동으로 수정해요.

수정된 프롬프트는 이미지 생성 호출의 revised_prompt 필드에서 확인할 수 있어요.

{
  "id": "ig_123",
  "type": "image_generation_call",
  "status": "completed",
  "revised_prompt": "A gray tabby cat hugging an otter. The otter is wearing an orange scarf. Both animals are cute and friendly, depicted in a warm, heartwarming style.",
  "result": "..."
}

프롬프트 팁

이미지 생성은 프롬프트에 draw나 edit 같은 용어를 쓰면 가장 잘 동작해요.

예를 들어 이미지를 결합하고 싶다면 combine이나 merge 대신 "첫 번째 이미지를 편집해서 두 번째 이미지의 이 요소를 추가해줘"처럼 말할 수 있어요.

다중 턴 편집 (Multi-turn editing)

이전 응답 또는 이미지 ID를 참조해서 이미지를 반복적으로 편집할 수 있어요. 이렇게 하면 대화 턴을 넘나들며 이미지를 다듬을 수 있어요.

이전 응답 ID 사용

다중 턴 이미지 생성

import OpenAI from "openai";
const openai = new OpenAI();

const response = await openai.responses.create({
  model: "gpt-6-astra",
  input:
    "Generate an image of gray tabby cat hugging an otter with an orange scarf",
  tools: [{ type: "image_generation", model: "gpt-image-2.5-sunburst" }],
});

const imageData = response.output
  .filter((output) => output.type === "image_generation_call")
  .map((output) => output.result);

if (imageData.length > 0) {
  const imageBase64 = imageData[0];
  const fs = await import("fs");
  fs.writeFileSync("cat_and_otter.png", Buffer.from(imageBase64, "base64"));
}

// Follow up

const response_fwup = await openai.responses.create({
  model: "gpt-6-astra",
  previous_response_id: response.id,
  input: "Now make it look realistic",
  tools: [{ type: "image_generation", model: "gpt-image-2.5-sunburst" }],
});

const imageData_fwup = response_fwup.output
  .filter((output) => output.type === "image_generation_call")
  .map((output) => output.result);

if (imageData_fwup.length > 0) {
  const imageBase64 = imageData_fwup[0];
  const fs = await import("fs");
  fs.writeFileSync(
    "cat_and_otter_realistic.png",
    Buffer.from(imageBase64, "base64")
  );
}

from openai import OpenAI
import base64

client = OpenAI()

response = client.responses.create(
    model="gpt-6-astra",
    input="Generate an image of gray tabby cat hugging an otter with an orange scarf",
    tools=[{"type": "image_generation", "model": "gpt-image-2.5-sunburst"}],
)

image_data = [
    output.result
    for output in response.output
    if output.type == "image_generation_call"
]

if image_data:
    image_base64 = image_data[0]

    with open("cat_and_otter.png", "wb") as f:
        f.write(base64.b64decode(image_base64))


# Follow up

response_fwup = client.responses.create(
    model="gpt-6-astra",
    previous_response_id=response.id,
    input="Now make it look realistic",
    tools=[{"type": "image_generation", "model": "gpt-image-2.5-sunburst"}],
)

image_data_fwup = [
    output.result
    for output in response_fwup.output
    if output.type == "image_generation_call"
]

if image_data_fwup:
    image_base64 = image_data_fwup[0]
    with open("cat_and_otter_realistic.png", "wb") as f:
        f.write(base64.b64decode(image_base64))

package main

import (
	"context"
	"encoding/base64"
	"os"

	"github.com/openai/openai-go/v3"
	"github.com/openai/openai-go/v3/responses"
)

func main() {
	client := openai.NewClient()
	first, err := client.Responses.New(context.Background(), responses.ResponseNewParams{
		Model: "gpt-6-astra",
		Input: responses.ResponseNewParamsInputUnion{
			OfString: openai.String("Generate an image of gray tabby cat hugging an otter with an orange scarf"),
		},
		Tools: []responses.ToolUnionParam{{OfImageGeneration: &responses.ToolImageGenerationParam{Model: "gpt-image-2.5-sunburst"}}},
	})
	if err != nil {
		panic(err)
	}
	saveFirstGeneratedImage(first, "cat_and_otter.png")

	followUp, err := client.Responses.New(context.Background(), responses.ResponseNewParams{
		Model:              "gpt-6-astra",
		PreviousResponseID: openai.String(first.ID),
		Input: responses.ResponseNewParamsInputUnion{
			OfString: openai.String("Now make it look realistic"),
		},
		Tools: []responses.ToolUnionParam{{OfImageGeneration: &responses.ToolImageGenerationParam{Model: "gpt-image-2.5-sunburst"}}},
	})
	if err != nil {
		panic(err)
	}
	saveFirstGeneratedImage(followUp, "cat_and_otter_realistic.png")
}

func saveFirstGeneratedImage(response *responses.Response, filename string) {
	for _, output := range response.Output {
		if output.Type != "image_generation_call" {
			continue
		}
		image, err := base64.StdEncoding.DecodeString(output.AsImageGenerationCall().Result)
		if err != nil {
			panic(err)
		}
		if err := os.WriteFile(filename, image, 0o600); err != nil {
			panic(err)
		}
		return
	}
	panic("response did not include an image generation call")
}

import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.models.responses.ResponseCreateParams;
import com.openai.models.responses.Tool;
import java.nio.file.Files;
import java.nio.file.Path;
import java.util.Base64;

var first =
    client
        .responses()
        .create(
            ResponseCreateParams.builder()
                .model("gpt-6-astra")
                .input(
                    "Generate an image of a gray tabby cat hugging an otter with an orange scarf.")
                .addTool(Tool.ImageGeneration.builder().build())
                .build());
var firstImage =
    first.output().stream()
        .flatMap(item -> item.imageGenerationCall().stream())
        .findFirst()
        .orElseThrow(() -> new IllegalStateException("No image generation call returned"));
Files.write(
    Path.of("cat_and_otter.png"),
    Base64.getDecoder()
        .decode(
            firstImage
                .result()
                .orElseThrow(() -> new IllegalStateException("No image returned"))));

var second =
    client
        .responses()
        .create(
            ResponseCreateParams.builder()
                .model("gpt-6-astra")
                .input("Now make it look realistic.")
                .previousResponseId(first.id())
                .addTool(Tool.ImageGeneration.builder().build())
                .build());
var secondImage =
    second.output().stream()
        .flatMap(item -> item.imageGenerationCall().stream())
        .findFirst()
        .orElseThrow(
            () -> new IllegalStateException("No follow-up image generation call returned"));
Files.write(
    Path.of("cat_and_otter_realistic.png"),
    Base64.getDecoder()
        .decode(
            secondImage
                .result()
                .orElseThrow(() -> new IllegalStateException("No follow-up image returned"))));

using OpenAI.Responses;
#pragma warning disable OPENAI001

string key = Environment.GetEnvironmentVariable("OPENAI_API_KEY")!;
ResponsesClient client = new(key);

CreateResponseOptions options = new() { Model = "gpt-6-astra" };
options.Tools.Add(ResponseTool.CreateImageGenerationTool(model: "gpt-image-2.5-sunburst"));
options.InputItems.Add(
    ResponseItem.CreateUserMessageItem(
        "Generate an image of a gray tabby cat hugging an otter with an orange scarf."
    )
);

ResponseResult first = await client.CreateResponseAsync(options);
ImageGenerationCallResponseItem initialImage = first
    .OutputItems.OfType<ImageGenerationCallResponseItem>()
    .First();
await File.WriteAllBytesAsync("cat_and_otter.png", initialImage.ImageResultBytes.ToArray());

CreateResponseOptions followUp = new()
{
    Model = "gpt-6-astra",
    PreviousResponseId = first.Id,
};
followUp.Tools.Add(ResponseTool.CreateImageGenerationTool(model: "gpt-image-2.5-sunburst"));
followUp.InputItems.Add(ResponseItem.CreateUserMessageItem("Now make it look realistic."));

ResponseResult second = await client.CreateResponseAsync(followUp);
ImageGenerationCallResponseItem updatedImage = second
    .OutputItems.OfType<ImageGenerationCallResponseItem>()
    .First();
await File.WriteAllBytesAsync(
    "cat_and_otter_realistic.png",
    updatedImage.ImageResultBytes.ToArray()
);

require "base64"
require "openai"

client = OpenAI::Client.new
first = client.responses.create(
  model: "gpt-6-astra",
  input: "Generate an image of a gray tabby cat hugging an otter with an orange scarf.",
  tools: [
    {
      type: :image_generation,
      model: "gpt-image-2.5-sunburst"
    }
  ]
)

first_image = first.output.find do |item|
  item.is_a?(OpenAI::Models::Responses::ResponseOutputItem::ImageGenerationCall)
end
unless first_image.is_a?(OpenAI::Models::Responses::ResponseOutputItem::ImageGenerationCall)
  raise "No image generation call returned"
end

encoded_image = first_image.result or raise "No image returned"
File.binwrite("cat_and_otter.png", Base64.strict_decode64(encoded_image))

follow_up = client.responses.create(
  model: "gpt-6-astra",
  input: "Now make it look realistic.",
  previous_response_id: first.id,
  tools: [
    {
      type: :image_generation,
      model: "gpt-image-2.5-sunburst"
    }
  ]
)

follow_up_image = follow_up.output.find do |item|
  item.is_a?(OpenAI::Models::Responses::ResponseOutputItem::ImageGenerationCall)
end
unless follow_up_image.is_a?(OpenAI::Models::Responses::ResponseOutputItem::ImageGenerationCall)
  raise "No follow-up image generation call returned"
end

encoded_image = follow_up_image.result or raise "No follow-up image returned"
File.binwrite("cat_and_otter_realistic.png", Base64.strict_decode64(encoded_image))

이미지 ID 사용

다중 턴 이미지 생성

import OpenAI from "openai";
const openai = new OpenAI();

const response = await openai.responses.create({
  model: "gpt-6-astra",
  input:
    "Generate an image of gray tabby cat hugging an otter with an orange scarf",
  tools: [{ type: "image_generation", model: "gpt-image-2.5-sunburst" }],
});

const imageGenerationCalls = response.output.filter(
  (output) => output.type === "image_generation_call"
);

const imageData = imageGenerationCalls.map((output) => output.result);

if (imageData.length > 0) {
  const imageBase64 = imageData[0];
  const fs = await import("fs");
  fs.writeFileSync("cat_and_otter.png", Buffer.from(imageBase64, "base64"));
}

// Follow up

const response_fwup = await openai.responses.create({
  model: "gpt-6-astra",
  input: [
    {
      role: "user",
      content: [{ type: "input_text", text: "Now make it look realistic" }],
    },
    {
      type: "image_generation_call",
      id: imageGenerationCalls[0].id,
    },
  ],
  tools: [{ type: "image_generation", model: "gpt-image-2.5-sunburst" }],
});

const imageData_fwup = response_fwup.output
  .filter((output) => output.type === "image_generation_call")
  .map((output) => output.result);

if (imageData_fwup.length > 0) {
  const imageBase64 = imageData_fwup[0];
  const fs = await import("fs");
  fs.writeFileSync(
    "cat_and_otter_realistic.png",
    Buffer.from(imageBase64, "base64")
  );
}

import openai
import base64

response = openai.responses.create(
    model="gpt-6-astra",
    input="Generate an image of gray tabby cat hugging an otter with an orange scarf",
    tools=[{"type": "image_generation", "model": "gpt-image-2.5-sunburst"}],
)

image_generation_calls = [
    output for output in response.output if output.type == "image_generation_call"
]

image_data = [output.result for output in image_generation_calls]

if image_data:
    image_base64 = image_data[0]

    with open("cat_and_otter.png", "wb") as f:
        f.write(base64.b64decode(image_base64))


# Follow up

response_fwup = openai.responses.create(
    model="gpt-6-astra",
    input=[
        {
            "role": "user",
            "content": [{"type": "input_text", "text": "Now make it look realistic"}],
        },
        {
            "type": "image_generation_call",
            "id": image_generation_calls[0].id,
        },
    ],
    tools=[{"type": "image_generation", "model": "gpt-image-2.5-sunburst"}],
)

image_data_fwup = [
    output.result
    for output in response_fwup.output
    if output.type == "image_generation_call"
]

if image_data_fwup:
    image_base64 = image_data_fwup[0]
    with open("cat_and_otter_realistic.png", "wb") as f:
        f.write(base64.b64decode(image_base64))

package main

import (
	"context"
	"encoding/base64"
	"encoding/json"
	"os"

	"github.com/openai/openai-go/v3"
	"github.com/openai/openai-go/v3/responses"
)

func main() {
	client := openai.NewClient()
	first, err := client.Responses.New(context.Background(), responses.ResponseNewParams{
		Model: "gpt-6-astra",
		Input: responses.ResponseNewParamsInputUnion{
			OfString: openai.String("Generate an image of gray tabby cat hugging an otter with an orange scarf"),
		},
		Tools: []responses.ToolUnionParam{{OfImageGeneration: &responses.ToolImageGenerationParam{Model: "gpt-image-2.5-sunburst"}}},
	})
	if err != nil {
		panic(err)
	}
	call := firstImageGenerationCall(first)
	saveImage("cat_and_otter.png", call.Result)
	input := outputAsInput(first.Output)
	input = append(input, responses.ResponseInputItemParamOfMessage(
		responses.ResponseInputMessageContentListParam{responses.ResponseInputContentParamOfInputText("Now make it look realistic")},
		responses.EasyInputMessageRoleUser,
	))

	followUp, err := client.Responses.New(context.Background(), responses.ResponseNewParams{
		Model: "gpt-6-astra",
		Input: responses.ResponseNewParamsInputUnion{OfInputItemList: input},
		Tools: []responses.ToolUnionParam{{OfImageGeneration: &responses.ToolImageGenerationParam{Model: "gpt-image-2.5-sunburst"}}},
	})
	if err != nil {
		panic(err)
	}
	saveImage("cat_and_otter_realistic.png", firstImageGenerationCall(followUp).Result)
}

func firstImageGenerationCall(response *responses.Response) responses.ResponseOutputItemImageGenerationCall {
	for _, output := range response.Output {
		if output.Type == "image_generation_call" {
			return output.AsImageGenerationCall()
		}
	}
	panic("response did not include an image generation call")
}

func outputAsInput(output []responses.ResponseOutputItemUnion) []responses.ResponseInputItemUnionParam {
	input := make([]responses.ResponseInputItemUnionParam, 0, len(output))
	for _, item := range output {
		var converted responses.ResponseInputItemUnion
		if err := json.Unmarshal([]byte(item.RawJSON()), &converted); err != nil {
			panic(err)
		}
		input = append(input, converted.ToParam())
	}
	return input
}

func saveImage(filename, encoded string) {
	image, err := base64.StdEncoding.DecodeString(encoded)
	if err != nil {
		panic(err)
	}
	if err := os.WriteFile(filename, image, 0o600); err != nil {
		panic(err)
	}
}

import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.core.JsonValue;
import com.openai.models.responses.ResponseCreateParams;
import com.openai.models.responses.ResponseInputItem;
import com.openai.models.responses.Tool;
import java.nio.file.Files;
import java.nio.file.Path;
import java.util.Base64;
import java.util.List;
import java.util.Map;

var first =
    client
        .responses()
        .create(
            ResponseCreateParams.builder()
                .model("gpt-6-astra")
                .input(
                    "Generate an image of a gray tabby cat hugging an otter with an orange scarf.")
                .addTool(Tool.ImageGeneration.builder().build())
                .build());
var firstImage =
    first.output().stream()
        .flatMap(item -> item.imageGenerationCall().stream())
        .findFirst()
        .orElseThrow(() -> new IllegalStateException("No image generation call returned"));
Files.write(
    Path.of("cat_and_otter.png"),
    Base64.getDecoder()
        .decode(
            firstImage
                .result()
                .orElseThrow(() -> new IllegalStateException("No image returned"))));

var second =
    client
        .responses()
        .create(
            ResponseCreateParams.builder()
                .model("gpt-6-astra")
                .inputOfResponse(
                    List.of(
                        ResponseInputItem.ofMessage(
                            ResponseInputItem.Message.builder()
                                .role(ResponseInputItem.Message.Role.USER)
                                .addInputTextContent("Now make it look realistic.")
                                .build()),
                        JsonValue.from(
                                Map.of("type", "image_generation_call", "id", firstImage.id()))
                            .convert(ResponseInputItem.class)))
                .addTool(Tool.ImageGeneration.builder().build())
                .build());
var secondImage =
    second.output().stream()
        .flatMap(item -> item.imageGenerationCall().stream())
        .findFirst()
        .orElseThrow(
            () -> new IllegalStateException("No follow-up image generation call returned"));
Files.write(
    Path.of("cat_and_otter_realistic.png"),
    Base64.getDecoder()
        .decode(
            secondImage
                .result()
                .orElseThrow(() -> new IllegalStateException("No follow-up image returned"))));

using OpenAI.Responses;
#pragma warning disable OPENAI001

string key = Environment.GetEnvironmentVariable("OPENAI_API_KEY")!;
ResponsesClient client = new(key);

CreateResponseOptions options = new() { Model = "gpt-6-astra" };
options.Tools.Add(ResponseTool.CreateImageGenerationTool(model: "gpt-image-2.5-sunburst"));
options.InputItems.Add(
    ResponseItem.CreateUserMessageItem(
        "Generate an image of a gray tabby cat hugging an otter with an orange scarf."
    )
);

ResponseResult first = await client.CreateResponseAsync(options);
ImageGenerationCallResponseItem initialImage = first
    .OutputItems.OfType<ImageGenerationCallResponseItem>()
    .First();
await File.WriteAllBytesAsync("cat_and_otter.png", initialImage.ImageResultBytes.ToArray());

CreateResponseOptions followUp = new() { Model = "gpt-6-astra" };
followUp.Tools.Add(ResponseTool.CreateImageGenerationTool(model: "gpt-image-2.5-sunburst"));
followUp.InputItems.Add(ResponseItem.CreateUserMessageItem("Now make it look realistic."));
followUp.InputItems.Add(ResponseItem.CreateReferenceItem(initialImage.Id));

ResponseResult second = await client.CreateResponseAsync(followUp);
ImageGenerationCallResponseItem updatedImage = second
    .OutputItems.OfType<ImageGenerationCallResponseItem>()
    .First();
await File.WriteAllBytesAsync(
    "cat_and_otter_realistic.png",
    updatedImage.ImageResultBytes.ToArray()
);

require "base64"
require "openai"

client = OpenAI::Client.new
first = client.responses.create(
  model: "gpt-6-astra",
  input: "Generate an image of a gray tabby cat hugging an otter with an orange scarf.",
  tools: [
    {
      type: :image_generation,
      model: "gpt-image-2.5-sunburst"
    }
  ]
)

first_image = first.output.find do |item|
  item.is_a?(OpenAI::Models::Responses::ResponseOutputItem::ImageGenerationCall)
end
unless first_image.is_a?(OpenAI::Models::Responses::ResponseOutputItem::ImageGenerationCall)
  raise "No image generation call returned"
end

encoded_image = first_image.result or raise "No image returned"
File.binwrite("cat_and_otter.png", Base64.strict_decode64(encoded_image))

follow_up = client.responses.create(
  model: "gpt-6-astra",
  input: [
    {
      role: :user,
      content: [
        {
          type: :input_text,
          text: "Now make it look realistic."
        }
      ]
    },
    {
      type: :image_generation_call,
      id: first_image.id
    }
  ],
  tools: [
    {
      type: :image_generation,
      model: "gpt-image-2.5-sunburst"
    }
  ]
)

follow_up_image = follow_up.output.find do |item|
  item.is_a?(OpenAI::Models::Responses::ResponseOutputItem::ImageGenerationCall)
end
unless follow_up_image.is_a?(OpenAI::Models::Responses::ResponseOutputItem::ImageGenerationCall)
  raise "No follow-up image generation call returned"
end

encoded_image = follow_up_image.result or raise "No follow-up image returned"
File.binwrite("cat_and_otter_realistic.png", Base64.strict_decode64(encoded_image))

스트리밍 (Streaming)

이미지 생성 도구는 최종 결과를 만드는 동안 부분 이미지(partial image)를 스트리밍할 수 있어요. 이렇게 하면 사용자에게 더 빠른 시각적 피드백을 제공하고 체감 지연 시간을 개선해요.

partial_images 파라미터로 부분 이미지 수(1-3)를 설정할 수 있어요.

이미지 스트리밍

import OpenAI from "openai";
import fs from "fs";
const openai = new OpenAI();

function saveBase64Image(filename, imageBase64) {
  const imageBuffer = Buffer.from(imageBase64, "base64");
  fs.writeFileSync(filename, imageBuffer);
}

const stream = await openai.responses.create({
  model: "gpt-6-astra",
  input:
    "Draw a gorgeous image of a river made of white owl feathers, snaking its way through a serene winter landscape",
  stream: true,
  tools: [
    { type: "image_generation", model: "gpt-image-2.5-sunburst", partial_images: 2 },
  ],
});

for await (const event of stream) {
  if (event.type === "response.image_generation_call.partial_image") {
    const idx = event.partial_image_index;
    saveBase64Image(`river-partial-${idx}.png`, event.partial_image_b64);
  } else if (event.type === "response.completed") {
    const imageData = event.response.output
      .filter((output) => output.type === "image_generation_call")
      .map((output) => output.result);

    if (imageData.length > 0) {
      saveBase64Image("river-final.png", imageData[0]);
    }
  }
}

from openai import OpenAI
import base64

client = OpenAI()


def save_base64_image(filename, image_base64):
    image_bytes = base64.b64decode(image_base64)
    with open(filename, "wb") as f:
        f.write(image_bytes)


stream = client.responses.create(
    model="gpt-6-astra",
    input="Draw a gorgeous image of a river made of white owl feathers, snaking its way through a serene winter landscape",
    stream=True,
    tools=[
        {"type": "image_generation", "model": "gpt-image-2.5-sunburst", "partial_images": 2}
    ],
)

for event in stream:
    if event.type == "response.image_generation_call.partial_image":
        idx = event.partial_image_index
        save_base64_image(f"river-partial-{idx}.png", event.partial_image_b64)
    elif event.type == "response.completed":
        image_data = [
            output.result
            for output in event.response.output
            if output.type == "image_generation_call"
        ]

        if image_data:
            save_base64_image("river-final.png", image_data[0])

package main

import (
	"context"
	"encoding/base64"
	"fmt"
	"os"

	"github.com/openai/openai-go/v3"
	"github.com/openai/openai-go/v3/responses"
)

func main() {
	client := openai.NewClient()
	stream := client.Responses.NewStreaming(context.Background(), responses.ResponseNewParams{
		Model: "gpt-6-astra",
		Input: responses.ResponseNewParamsInputUnion{
			OfString: openai.String("Draw a gorgeous image of a river made of white owl feathers, snaking its way through a serene winter landscape"),
		},
		Tools: []responses.ToolUnionParam{{OfImageGeneration: &responses.ToolImageGenerationParam{Model: "gpt-image-2.5-sunburst", PartialImages: openai.Int(2)}}},
	})
	for stream.Next() {
		event := stream.Current()
		if event.Type == "response.image_generation_call.partial_image" {
			partial := event.AsResponseImageGenerationCallPartialImage()
			saveImage(fmt.Sprintf("river-partial-%d.png", partial.PartialImageIndex), partial.PartialImageB64)
		}
		if event.Type == "response.completed" {
			for _, output := range event.AsResponseCompleted().Response.Output {
				if output.Type == "image_generation_call" {
					saveImage("river-final.png", output.AsImageGenerationCall().Result)
				}
			}
		}
	}
	if err := stream.Err(); err != nil {
		panic(err)
	}
}

func saveImage(filename, encoded string) {
	image, err := base64.StdEncoding.DecodeString(encoded)
	if err != nil {
		panic(err)
	}
	if err := os.WriteFile(filename, image, 0o600); err != nil {
		panic(err)
	}
}

import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.core.http.StreamResponse;
import com.openai.models.responses.ResponseCreateParams;
import com.openai.models.responses.ResponseStreamEvent;
import com.openai.models.responses.Tool;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Path;
import java.util.Base64;

ResponseCreateParams params =
    ResponseCreateParams.builder()
        .model("gpt-6-astra")
        .input("Generate an image of a river made of white owl feathers.")
        .addTool(Tool.ImageGeneration.builder().partialImages(2).build())
        .build();

try (StreamResponse<ResponseStreamEvent> stream = client.responses().createStreaming(params)) {
  var events = stream.stream().iterator();
  while (events.hasNext()) {
    ResponseStreamEvent event = events.next();
    if (event.imageGenerationCallPartialImage().isPresent()) {
      var partial = event.imageGenerationCallPartialImage().orElseThrow();
      Files.write(
          Path.of("river-partial-" + partial.partialImageIndex() + ".png"),
          Base64.getDecoder().decode(partial.partialImageB64()));
    }
    if (event.completed().isPresent()) {
      var image =
          event.completed().orElseThrow().response().output().stream()
              .flatMap(item -> item.imageGenerationCall().stream())
              .findFirst()
              .orElseThrow(() -> new IllegalStateException("No generated image returned"));
      Files.write(
          Path.of("river-final.png"),
          Base64.getDecoder()
              .decode(
                  image
                      .result()
                      .orElseThrow(
                          () -> new IllegalStateException("No final image returned"))));
    }
  }
}

require "base64"
require "openai"

client = OpenAI::Client.new
stream = client.responses.stream(
  model: "gpt-6-astra",
  input: "Generate an image of a river made of white owl feathers.",
  tools: [
    {
      type: :image_generation,
      model: "gpt-image-2.5-sunburst",
      partial_images: 2
    }
  ]
)

stream.each do |event|
  case event
  when OpenAI::Models::Responses::ResponseImageGenCallPartialImageEvent
    image = Base64.strict_decode64(event.partial_image_b64)
    File.binwrite("river-partial-#{event.partial_image_index}.png", image)
  when OpenAI::Models::Responses::ResponseCompletedEvent
    image_call = event.response.output.find do |item|
      item.is_a?(OpenAI::Models::Responses::ResponseOutputItem::ImageGenerationCall)
    end
    next unless image_call.is_a?(OpenAI::Models::Responses::ResponseOutputItem::ImageGenerationCall)

    File.binwrite(
      "river-final.png",
      Base64.strict_decode64(image_call.result)
    )
  end
end

지원되는 모델

다음 모델이 이미지 생성 도구를 지원해요.

• gpt-5.5
• gpt-5.4-mini
• gpt-5.4-nano
• gpt-5.2
• gpt-5
• gpt-5-nano
• o3
• gpt-4.1
• gpt-4.1-mini
• gpt-4.1-nano
• gpt-4o
• gpt-4o-mini

더 알아보기 (Learn more)

• 이미지 생성 가이드에서 이미지 생성·편집과 출력 커스터마이징을 확인하세요.
• Responses API의 기타 내장 도구에 대해 더 알아보세요.