Geração de imagens

A ferramenta de geração de imagens permite gerar imagens usando um prompt de texto e, opcionalmente, imagens de entrada. Ela usa modelos GPT Image, incluindo gpt-image-2.5-sunburst, gpt-image-2.5-flare, gpt-image-2, gpt-image-1.5, gpt-image-1 e gpt-image-1-mini, e otimiza automaticamente as entradas de texto para melhorar o desempenho.

Defina model da ferramenta image_generation como gpt-image-2.5-sunburst para edições precisas ou como gpt-image-2.5-flare para gerar imagens rapidamente e com alta qualidade. Use um modelo principal compatível no campo model de nível superior da API Responses.

Para saber mais sobre geração de imagens, consulte nosso guia de geração de imagens.

Uso

Ao incluir a ferramenta image_generation na sua solicitação, o modelo pode decidir quando e como gerar imagens durante a conversa, usando seu prompt e as imagens de entrada fornecidas.

O resultado da chamada de ferramenta image_generation_call incluirá uma imagem codificada em base64.

Gerar uma imagem

Python

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20import 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"));
}

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22from 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))

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42package 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")
}

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20using 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());

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24require "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))

Você pode fornecer imagens de entrada usando IDs de arquivos ou dados em base64.

Para forçar a chamada da ferramenta de geração de imagens, você pode definir o parâmetro tool_choice como {"type": "image_generation"}.

Opções da ferramenta

Você pode configurar as seguintes opções de saída como parâmetros da ferramenta de geração de imagens:

Tamanho: dimensões da imagem, por exemplo, 1024 × 1024 ou 1024 × 1536
Qualidade: qualidade de renderização, por exemplo, baixa, média ou alta
Formato: formato do arquivo de saída
Compressão: nível de compressão (0-100%) para os formatos JPEG e WebP
Fundo: transparente, opaco ou automático
Ação: define se a solicitação deve escolher automaticamente a ação, gerar uma imagem ou editar uma imagem

size, quality e background aceitam a opção auto, com a qual o modelo seleciona automaticamente a melhor opção com base no prompt.

Para gpt-image-2.5-sunburst e gpt-image-2.5-flare, quality também aceita xhigh e max. Esses valores não são compatíveis com modelos GPT Image anteriores. A qualidade padrão continua sendo auto.

O gpt-image-2 aceita valores flexíveis de size que atendam às suas restrições de resolução. Fundos transparentes estão disponíveis em versão prévia; configure background: "transparent" para solicitar um. Use png (o padrão) ou webp; jpeg não é compatível com fundos transparentes.

Para mais detalhes sobre as opções disponíveis, consulte o guia de geração de imagens.

Ao usar a ferramenta de geração de imagens da API Responses, os modelos GPT Image compatíveis podem escolher entre gerar uma nova imagem ou editar uma imagem já presente na conversa. O parâmetro opcional action controla esse comportamento: mantenha action definido como auto para que o modelo escolha entre gerar ou editar, ou defina-o como generate ou edit para forçar o comportamento correspondente. Se não for especificado, o padrão será auto.

Prompt revisado

Ao usar a ferramenta de geração de imagens, o modelo principal, por exemplo, gpt-5.5, revisará automaticamente seu prompt para melhorar o desempenho.

Você pode acessar o prompt revisado no campo revised_prompt da chamada de geração de imagens:

1234567{
  "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": "..."
}

Dicas para criação de prompts

A geração de imagens funciona melhor quando você usa termos como draw ou edit no prompt.

Por exemplo, se quiser combinar imagens, em vez de dizer combine ou merge, você pode dizer algo como "edite a primeira imagem adicionando este elemento da segunda imagem".

Edição em múltiplas interações

Você pode editar imagens de forma iterativa referenciando IDs de respostas ou imagens anteriores. Isso permite refinar as imagens ao longo das interações na conversa.

Geração de imagens em múltiplas interações

Python

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41import 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")
  );
}

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43from 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))

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55package 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")
}

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36using 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()
);

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46require "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))

Geração de imagens em múltiplas interações

Python

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51import 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")
  );
}

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49import 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))

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73package 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)
	}
}

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33using 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()
);

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59require "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))

Streaming

A ferramenta de geração de imagens permite transmitir imagens parciais por streaming enquanto gera o resultado final. Isso oferece um retorno visual mais rápido aos usuários e reduz a latência percebida.

Você pode definir o número de imagens parciais (1-3) com o parâmetro partial_images.

Transmitir uma imagem por streaming

Python

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33import 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]);
    }
  }
}

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34from 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])

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49package 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)
	}
}

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33require "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

Modelos compatíveis

Os seguintes modelos são compatíveis com a ferramenta de geração de imagens:

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

Sugestões

Primeiros passos

Conceitos fundamentais

SDKs e CLI

Recursos

APIs legadas

Escolha um modelo

Texto e código

Criação de prompts

Raciocínio

Imagens e vídeo

Tempo real e áudio

Modelos especializados

API de Agentes

SDK de Agentes

ChatKit

Pesquisa e recuperação

Conecte ferramentas e dados

Crie fluxos de trabalho com ferramentas

Computador e código

Mídia

GPT-Live

Realtime API

Desenvolva com voz

Conexões

Processamento de áudio

Entre em produção

Desempenho e qualidade

Custo e taxa de processamento

Segurança e governança

Infraestrutura e acesso

Operações

Conceitos fundamentais

Plan

Build

Test and publish

Conversion specs

Guias

Recursos

Primeiros passos

Guias

File Upload

API

Measurement

Advertiser API

Referência da API

Primeiros passos

Fundamentos

Explorar

Disponível em

Versões

Fluxos de trabalho

Capacidades

Referência

Personalização

Arquivo de configuração

Configuração de agentes

Amplie os recursos do ChatGPT e do Codex

Linux

Windows

Fluxos de trabalho de desenvolvimento

Amplie e automatize

Ambientes

Desenvolva com o Codex

Integrações de terceiros

Referência

Permissões

Codex Security

Segurança cibernética

Primeiros passos

ChatGPT Work

Identidade e autenticação

Acesso, política e modelos do workspace

Controles de plug-ins e conectores

Uso, governança e conformidade

Implantação e provedores de modelos

Comunidade

Blog

Comunidade

Blog