您可以通过 API 使用 gpt-image-2.5-sunburst 和 gpt-image-2.5-flare,根据文本提示词生成和编辑图像。如果工作流程最看重编辑精度,请选择 Sunburst。如果日常使用时需要快速生成高质量图像,请选择 Flare。您可以通过以下两种 API 使用图像生成功能:
使用 Image API 时,直接将 model 设置为 gpt-image-2.5-sunburst 或 gpt-image-2.5-flare。使用 Responses API 时,在请求的顶层选择受支持的主模型,并将 gpt-image-2.5-sunburst 或 gpt-image-2.5-flare 填入图像生成工具的 model 字段。
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19import OpenAI from "openai";import fs from "fs";const openai = new OpenAI();const prompt = `A children's book drawing of a veterinarian using a stethoscope tolisten to the heartbeat of a baby otter.`;const result = await openai.images.generate({ model: "gpt-image-2.5-sunburst", prompt,});// Save the image to a fileconst image_base64 = result.data?.[0]?.b64_json;if (!image_base64) throw new Error("Image generation returned no image.");const image_bytes = Buffer.from(image_base64, "base64");fs.writeFileSync("otter.png", image_bytes);
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18from openai import OpenAIimport base64client = OpenAI()prompt ="""A children's book drawing of a veterinarian using a stethoscope tolisten to the heartbeat of a baby otter."""result = client.images.generate(model="gpt-image-2.5-sunburst", prompt=prompt)image_base64 = result.data[0].b64_jsonimage_bytes = base64.b64decode(image_base64)# Save the image to a filewithopen("otter.png", "wb") as f: f.write(image_bytes)
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
28package mainimport ( "context" "encoding/base64" "os" "github.com/openai/openai-go/v3")func main() { client := openai.NewClient() result, err := client.Images.Generate(context.Background(), openai.ImageGenerateParams{ Model: openai.ImageModel("gpt-image-2.5-sunburst"), Prompt: "A children's book drawing of a veterinarian using a stethoscope to " + "listen to the heartbeat of a baby otter.", }) if err != nil { panic(err) } image, err := base64.StdEncoding.DecodeString(result.Data[0].B64JSON) if err != nil { panic(err) } if err := os.WriteFile("otter.png", image, 0o600); err != nil { panic(err) }}
1
2
3
4
5
6
7
8
9
10
11
12using OpenAI.Images;string key = Environment.GetEnvironmentVariable("OPENAI_API_KEY")!;string model = "gpt-image-2.5-sunburst";ImageClient client = new(model, key);GeneratedImage image = await client.GenerateImageAsync( "A children's book drawing of a veterinarian using a stethoscope to " + "listen to the heartbeat of a baby otter.");await File.WriteAllBytesAsync("otter.png", image.ImageBytes.ToArray());
1
2
3
4
5
6
7
8
9
10
11
12
13require "base64"require "openai"client = OpenAI::Client.newresult = client.images.generate( model: "gpt-image-2.5-sunburst", prompt: "A watercolor robot reading in a library")generated_image = result.data&.first or raise "No image returned"File.binwrite( "generated-image.png", Base64.strict_decode64(generated_image.b64_json))
1
2
3
4
5
6
7curl -X POST "https://api.openai.com/v1/images/generations" \ -H "Authorization: Bearer $OPENAI_API_KEY" \ -H "Content-type: application/json" \ -d '{ "model": "gpt-image-2.5-sunburst", "prompt": "A children'\''s book drawing of a veterinarian using a stethoscope to listen to the heartbeat of a baby otter." }' | jq -r '.data[0].b64_json' | base64 --decode > otter.png
1
2
3
4
5openai images generate \ --model gpt-image-2.5-sunburst \ --prompt "A children's book drawing of a veterinarian using a stethoscope to listen to the heartbeat of a baby otter." \ --raw-output \ --transform 'data.0.b64_json' | base64 --decode > otter.png
Responses API
生成图像
Python
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21import OpenAI from "openai";const openai = new OpenAI();const response = await openai.responses.create({ model: "gpt-6.2-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 fileconst imageData = response.output .filter((output) => output.type === "image_generation_call") .map((output) => output.result);if (imageData.length > 0) { const imageBase64 = imageData[0]; if (!imageBase64) throw new Error("Image generation returned no image."); 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 OpenAIimport base64client = OpenAI()response = client.responses.create(model="gpt-6.2-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 fileimage_data = [ output.resultfor output in response.outputif output.type =="image_generation_call"]if image_data: image_base64 = image_data[0]withopen("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 mainimport ( "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-6.2-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 OPENAI001string key = Environment.GetEnvironmentVariable("OPENAI_API_KEY")!;ResponsesClient client = new(key);CreateResponseOptions options = new() { Model = "gpt-6.2-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.newresponse = client.responses.create( model: "gpt-6.2-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)endunless image_call.is_a?(OpenAI::Models::Responses::ResponseOutputItem::ImageGenerationCall) raise "No image generation call returned"endencoded_image = image_call.result or raise "No image returned"File.binwrite("otter.png", Base64.strict_decode64(encoded_image))
多轮图像生成
使用 Responses API 时,您可以在上下文中提供图像生成调用的输出(也可以只提供图像 ID),或者使用 previous_response_id 参数,以此构建包含图像生成的多轮对话。
随着对话推进,您可以逐轮调整提示词、给出新指令,逐步修改图像效果。
Responses API 和 Image API 都支持流式图像生成。您可以在 API 生成图像的过程中,以流式方式接收中间图像,让用户实时看到生成进度。
您可以调整 partial_images 参数,接收 0 至 3 张中间图像。
如果将 partial_images 设置为 0,您只会收到最终图像。
如果设置的值大于 0,而完整图像生成得较快,您收到的中间图像可能少于请求的数量。
Responses API
流式接收图像
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
34import OpenAI from "openai";import fs from "fs";const openai = new OpenAI();function saveBase64Image(filename: string, imageBase64: string | null) { if (!imageBase64) throw new Error("Image generation returned no image."); const imageBuffer = Buffer.from(imageBase64, "base64"); fs.writeFileSync(filename, imageBuffer);}const stream = await openai.responses.create({ model: "gpt-6.2-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 OpenAIimport base64client = OpenAI()defsave_base64_image(filename, image_base64): image_bytes = base64.b64decode(image_base64)withopen(filename, "wb") as f: f.write(image_bytes)stream = client.responses.create(model="gpt-6.2-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.resultfor output in event.response.outputif 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 mainimport ( "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-6.2-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.newstream = client.responses.stream( model: "gpt-6.2-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) ) endend
Image API
流式接收图像
Python
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23import fs from "fs";import OpenAI from "openai";const openai = new OpenAI();const prompt = "Draw a gorgeous image of a river made of white owl feathers, snaking its way through a serene winter landscape";const stream = await openai.images.generate({ prompt: prompt, model: "gpt-image-2.5-sunburst", stream: true, partial_images: 2,});for await (const event of stream) { if (event.type === "image_generation.partial_image") { const idx = event.partial_image_index; const imageBase64 = event.b64_json; if (!imageBase64) throw new Error("Image generation returned no image."); const imageBuffer = Buffer.from(imageBase64, "base64"); fs.writeFileSync(`river${idx}.png`, imageBuffer); }}
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19from openai import OpenAIimport base64client = OpenAI()stream = client.images.generate(prompt="Draw a gorgeous image of a river made of white owl feathers, snaking its way through a serene winter landscape",model="gpt-image-2.5-sunburst",stream=True,partial_images=2,)for event in stream:if event.type =="image_generation.partial_image": idx = event.partial_image_index image_base64 = event.b64_json image_bytes = base64.b64decode(image_base64)withopen(f"river{idx}.png", "wb") as f: f.write(image_bytes)
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
40package mainimport ( "context" "encoding/base64" "fmt" "os" "github.com/openai/openai-go/v3")func main() { client := openai.NewClient() stream := client.Images.GenerateStreaming(context.Background(), openai.ImageGenerateParams{ Model: openai.ImageModel("gpt-image-2.5-sunburst"), Prompt: "Draw a gorgeous image of a river made of white owl feathers, snaking its way through a serene winter landscape", PartialImages: openai.Int(2), }) for stream.Next() { event := stream.Current() if event.Type != "image_generation.partial_image" { continue } partial := event.AsImageGenerationPartialImage() saveImage(fmt.Sprintf("river%d.png", partial.PartialImageIndex), partial.B64JSON) } 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
25import com.openai.client.okhttp.OpenAIOkHttpClient;import com.openai.models.images.ImageGenerateParams;import java.nio.file.Files;import java.nio.file.Path;import java.util.Base64;var params = ImageGenerateParams.builder() .prompt( "Draw a gorgeous image of a river made of white owl feathers, snaking its way through a serene winter landscape") .model("gpt-image-2.5-sunburst") .partialImages(2) .build();try (var stream = client.images().generateStreaming(params)) { var events = stream.stream().iterator(); while (events.hasNext()) { var event = events.next(); if (event.generationPartialImage().isPresent()) { var partial = event.generationPartialImage().orElseThrow(); Files.write( Path.of("river" + partial.partialImageIndex() + ".png"), Base64.getDecoder().decode(partial.b64Json())); } }}
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16require "base64"require "openai"client = OpenAI::Client.newstream = client.images.generate_stream_raw( model: "gpt-image-2.5-sunburst", prompt: "A river made of white owl feathers in a winter landscape", partial_images: 2)stream.each do |event| next unless event.is_a?(OpenAI::Models::ImageGenPartialImageEvent) image = Base64.strict_decode64(event.b64_json) File.binwrite("river#{event.partial_image_index}.png", image)end
结果
中间图像 1
中间图像 2
最终图像
提示词:画一幅美丽的冬日风景,一条由白色猫头鹰羽毛汇成的河流蜿蜒穿过宁静的原野。
改写后的提示词
在 Responses API 中使用图像生成工具时,主模型(例如 gpt-5.5)会自动改写您的提示词,以改善生成效果。
您可以在图像生成调用的 revised_prompt 字段中查看改写后的提示词:
包含改写后提示词的响应
1
2
3
4
5
6
7{"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": "..."}
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
58require "base64"require "openai"require "pathname"client = OpenAI::Client.newbase64_images = ["body-lotion.png", "soap.png"].map do |path| Base64.strict_encode64(File.binread(path))endfile_ids = [ client.files.create(file: Pathname("bath-bomb.png"), purpose: :vision).id, client.files.create(file: Pathname("incense-kit.png"), purpose: :vision).id]prompt = <<~PROMPT Generate a photorealistic image of a gift basket on a white background labeled 'Relax & Unwind' with a ribbon and handwriting-like font, containing all the items in the reference pictures.PROMPTresponse = client.responses.create( model: "gpt-6.2-astra", input: [ { role: :user, content: [ { type: :input_text, text: prompt }, *base64_images.map do |image| { type: :input_image, image_url: "data:image/png;base64,#{image}" } end, *file_ids.map do |file_id| { type: :input_image, file_id: file_id } end ] } ], 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)endunless image_call.is_a?(OpenAI::Models::Responses::ResponseOutputItem::ImageGenerationCall) raise "No image generation call returned"endFile.binwrite("gift-basket.png", Base64.strict_decode64(image_call.result))
Image API
编辑图像
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
38import fs from "fs";import OpenAI, { toFile } from "openai";const client = new OpenAI();const prompt = `Generate a photorealistic image of a gift basket on a white backgroundlabeled 'Relax & Unwind' with a ribbon and handwriting-like font,containing all the items in the reference pictures.`;const imageFiles = [ "fixtures/bath-bomb.png", "fixtures/body-lotion.png", "fixtures/incense-kit.png", "fixtures/soap.png",];const images = await Promise.all( imageFiles.map( async (file) => await toFile(fs.createReadStream(file), null, { type: "image/png", }) ));const response = await client.images.edit({ model: "gpt-image-2.5-sunburst", image: images, prompt,});// Save the image to a fileconst image_base64 = response.data?.[0]?.b64_json;if (!image_base64) throw new Error("Image generation returned no image.");const image_bytes = Buffer.from(image_base64, "base64");fs.writeFileSync("basket.png", image_bytes);
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
28import base64from openai import OpenAIclient = OpenAI()prompt ="""Generate a photorealistic image of a gift basket on a white backgroundlabeled 'Relax & Unwind' with a ribbon and handwriting-like font,containing all the items in the reference pictures."""result = client.images.edit(model="gpt-image-2.5-sunburst",image=[open("body-lotion.png", "rb"),open("bath-bomb.png", "rb"),open("incense-kit.png", "rb"),open("soap.png", "rb"), ],prompt=prompt,)image_base64 = result.data[0].b64_jsonimage_bytes = base64.b64decode(image_base64)# Save the image to a filewithopen("gift-basket.png", "wb") as f: f.write(image_bytes)
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
65package mainimport ( "context" "encoding/base64" "io" "os" "github.com/openai/openai-go/v3")func main() { client := openai.NewClient() files, closeFiles := openImages( "bath-bomb.png", "body-lotion.png", "incense-kit.png", "soap.png", ) defer closeFiles() response, err := client.Images.Edit(context.Background(), openai.ImageEditParams{ Model: openai.ImageModel("gpt-image-2.5-sunburst"), Image: openai.ImageEditParamsImageUnion{OfFileArray: files}, Prompt: "Generate a photorealistic image of a gift basket on a white background " + "labeled 'Relax & Unwind' with a ribbon and handwriting-like font, containing all the items in the reference pictures.", }) if err != nil { panic(err) } saveImage("basket.png", response.Data[0].B64JSON)}func openImages(names ...string) ([]io.Reader, func()) { images := make([]io.Reader, 0, len(names)) files := make([]*os.File, 0, len(names)) for _, name := range names { file, err := os.Open(name) if err != nil { closeFiles(files) panic(err) } images = append(images, openai.File(file, name, "image/png")) files = append(files, file) } return images, func() { closeFiles(files) }}func closeFiles(files []*os.File) { for _, file := range files { if err := file.Close(); 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
33
34
35
36
37
38
39
40
41
42
43
44
45import com.openai.client.OpenAIClient;import com.openai.client.okhttp.OpenAIOkHttpClient;import com.openai.core.MultipartField;import com.openai.models.images.ImageEditParams;import java.io.IOException;import java.io.InputStream;import java.nio.file.Files;import java.nio.file.Path;import java.util.Base64;import java.util.List;Path lotion = Path.of(System.getenv("OPENAI_EXAMPLE_IMAGE_PATH"));Path soap = Path.of(System.getenv("OPENAI_EXAMPLE_IMAGE_PATH_2"));Path bathBomb = Path.of(System.getenv("OPENAI_EXAMPLE_IMAGE_PATH_3"));Path incense = Path.of(System.getenv("OPENAI_EXAMPLE_IMAGE_PATH_4"));try (InputStream lotionImage = Files.newInputStream(lotion); InputStream bathBombImage = Files.newInputStream(bathBomb); InputStream incenseImage = Files.newInputStream(incense); InputStream soapImage = Files.newInputStream(soap)) { var images = client .images() .edit( ImageEditParams.builder() .model("gpt-image-2.5-sunburst") .image( MultipartField.<ImageEditParams.Image>builder() .value( ImageEditParams.Image.ofInputStreams( List.of(lotionImage, bathBombImage, incenseImage, soapImage))) .contentType("image/png") .filename("gift-basket-reference.png") .build()) .prompt( """ Generate a photorealistic image of a gift basket on a white background labeled 'Relax & Unwind' with a ribbon and handwriting-like font, containing all the items in the reference pictures. """) .build()); Files.write( Path.of("gift-basket.png"), Base64.getDecoder().decode(images.data().orElseThrow().get(0).b64Json().orElseThrow()));}
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19require "base64"require "openai"require "pathname"client = OpenAI::Client.newimages = %w[body-lotion.png bath-bomb.png incense-kit.png soap.png].map do |path| OpenAI::FilePart.new(Pathname(path), content_type: "image/png")endresult = client.images.edit( image: images, model: "gpt-image-2.5-sunburst", prompt: <<~PROMPT Generate a photorealistic image of a gift basket on a white background labeled 'Relax & Unwind' with a ribbon and handwriting-like font, containing all the items in the reference pictures. PROMPT)generated_image = result.data&.first or raise "No image returned"File.binwrite("gift-basket.png", Base64.strict_decode64(generated_image.b64_json))
1
2
3
4
5
6
7
8
9
10curl -s -D >(grep -i x-request-id >&2) \ -o >(jq -r '.data[0].b64_json' | base64 --decode > gift-basket.png) \ -X POST "https://api.openai.com/v1/images/edits" \ -H "Authorization: Bearer $OPENAI_API_KEY" \ -F "model=gpt-image-2.5-sunburst" \ -F "image[]=@body-lotion.png" \ -F "image[]=@bath-bomb.png" \ -F "image[]=@incense-kit.png" \ -F "image[]=@soap.png" \ -F 'prompt=Generate a photorealistic image of a gift basket on a white background labeled "Relax & Unwind" with a ribbon and handwriting-like font, containing all the items in the reference pictures'
1
2
3
4
5
6
7
8
9openai images edit \ --model gpt-image-2.5-sunburst \ --image body-lotion.png \ --image bath-bomb.png \ --image incense-kit.png \ --image soap.png \ --prompt 'Generate a photorealistic image of a gift basket on a white background labeled "Relax & Unwind" with a ribbon and handwriting-like font, containing all the items in the reference pictures' \ --raw-output \ --transform 'data.0.b64_json' | base64 --decode > gift-basket.png
使用标注引导编辑
使用支持标注引导编辑的 GPT-Image-2.5 或 GPT-Image-2.6 模型进行此类编辑时,请使用 Image API 的 POST /v1/images/edits 端点。如果使用 Responses API 进行局部编辑,请使用蒙版。用标注图层指出需要修改的位置,并在 prompt 中说明要如何修改。
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21import base64from pathlib import Pathfrom openai import OpenAIclient = OpenAI()overlay = Path("markup-vase-overlay-gpt-image-2-6.png")withopen("kitchen.webp", "rb") as image, overlay.open("rb") as markup: result = client.images.edit(model="gpt-image-2.6-flare",image=("kitchen.webp", image, "image/webp"),markup=(overlay.name, markup, "image/png"),quality="medium",prompt=("Recolor the striped vase inside the red circle to solid cobalt blue ceramic. ""Keep its shape, the dried stems, and the rest of the kitchen unchanged. ""Remove the red circle from the finished photo." ), )Path("vase.png").write_bytes(base64.b64decode(result.data[0].b64_json))
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
31client := openai.NewClient()image, err := os.Open("kitchen.webp")if err != nil { panic(err)}defer image.Close()markup, err := os.Open("markup-vase-overlay-gpt-image-2-6.png")if err != nil { panic(err)}defer markup.Close()response, err := client.Images.Edit(context.Background(), openai.ImageEditParams{ Model: openai.ImageModel("gpt-image-2.6-flare"), Image: openai.ImageEditParamsImageUnion{OfFile: openai.File(image, "kitchen.webp", "image/webp")}, Markup: openai.File(markup, "markup-vase-overlay-gpt-image-2-6.png", "image/png"), Quality: "medium", Prompt: "Recolor the striped vase inside the red circle to solid cobalt blue ceramic. " + "Keep its shape, the dried stems, and the rest of the kitchen unchanged. " + "Remove the red circle from the finished photo.",})if err != nil { panic(err)}result, err := base64.StdEncoding.DecodeString(response.Data[0].B64JSON)if err != nil { panic(err)}if err := os.WriteFile("vase.png", result, 0o600); err != nil { panic(err)}
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17require "openai"require "pathname"require "base64"client = OpenAI::Client.newresult = client.images.edit( model: "gpt-image-2.6-flare", image: OpenAI::FilePart.new(Pathname("kitchen.webp"), content_type: "image/webp"), markup: OpenAI::FilePart.new(Pathname("markup-vase-overlay-gpt-image-2-6.png"), content_type: "image/png"), quality: :medium, prompt: "Recolor the striped vase inside the red circle to solid cobalt blue ceramic. " \ "Keep its shape, the dried stems, and the rest of the kitchen unchanged. " \ "Remove the red circle from the finished photo.")image = result.data&.first or raise "No image returned"encoded = image.b64_json or raise "No image data returned"File.binwrite("vase.png", Base64.strict_decode64(encoded))
1
2
3
4
5
6
7
8
9
10set -euo pipefailcurl --fail --silent --show-error https://api.openai.com/v1/images/edits \ -H "Authorization: Bearer $OPENAI_API_KEY" \ -F model=gpt-image-2.6-flare \ -F quality=medium \ -F 'image=@kitchen.webp;type=image/webp' \ -F 'markup=@markup-vase-overlay-gpt-image-2-6.png;type=image/png' \ -F 'prompt=Recolor the striped vase inside the red circle to solid cobalt blue ceramic. Keep its shape, the dried stems, and the rest of the kitchen unchanged. Remove the red circle from the finished photo.' | jq -er '.data[0].b64_json' | base64 --decode > vase.png
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21fromPILimport Imagefrom io import BytesIO# 1. Load your black & white mask as a grayscale imagemask = Image.open("mask.png").convert("L")# 2. Convert it to RGBA so it has space for an alpha channelmask_rgba = mask.convert("RGBA")# 3. Then use the mask itself to fill that alpha channelmask_rgba.putalpha(mask)# 4. Convert the mask into bytesbuf = BytesIO()mask_rgba.save(buf, format="PNG")mask_bytes = buf.getvalue()# 5. Save the resulting fileimg_path_mask_alpha ="mask_alpha.png"withopen(img_path_mask_alpha, "wb") as f: f.write(mask_bytes)
要输出 GIF,请设置 output_format: "gif",并将请求发送到 POST /v1/images/generations 或 POST /v1/images/edits。在提示词中描述动画主体、动作以及希望动画如何循环播放。如果要用静态参考图像引导动画生成,请将这些图像发送到图像编辑端点。
无论使用哪个端点,都可以通过以下请求字段配置 GIF 输出:
字段
取值
默认值
frame_count
2 至 16 帧
9
frame_duration_ms
每帧 20 至 2000 毫秒,按最接近的 10 毫秒取整
200
size
auto,或使用 WIDTHxHEIGHT 指定尺寸,宽和高都必须是 16 的正整数倍
auto
size 指定输出动画中每一帧的尺寸。帧数、每帧尺寸和动画数量必须同时满足模型的尺寸限制和单次请求的总像素上限。如果请求超出限制,请减少 frame_count、缩小每帧尺寸或减小 n。使用 auto 可让 API 在模型允许的范围内选择每帧尺寸。
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";constopenai=newOpenAI();try {// The same error handling pattern applies to image generation requests,// image edits, and Responses API tool calls that generate images.await openai.images.generate({ model: "gpt-image-2.5-sunburst", prompt: "Create a poster humiliating my coworker with insulting captions", });} catch (error) {if (!(error instanceofOpenAI.APIError) || error.code !=="moderation_blocked" ) {throw error; }constmoderationDetails= error.error &&"moderation_details"in error.error? (error.error.moderation_details as {categories?:string[];moderation_stage?:string; }):undefined;constcategories= moderationDetails?.categories ?? [];conststage= moderationDetails?.moderation_stage;let hint ="This request could not be completed because it did not meet safety requirements.";if (categories.includes("harassment")) { hint ="Try removing abusive or targeting language and focus on neutral visual details instead."; } elseif (stage ==="input") { hint ="Try revising the prompt or input images and submit the request again."; } elseif (stage ==="output") { hint ="The generated result was blocked by a safety check. Try changing the prompt and generating again."; } console.error("Image generation blocked", { request_id: error?.requestID, code: error?.code, moderation_details: moderationDetails, }); console.log(hint);}
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
40import openaifrom openai import OpenAIclient = OpenAI()try: # The same error handling pattern applies to image generation requests, # image edits, and Responses API tool calls that generate images. client.images.generate( model="gpt-image-2.5-sunburst", prompt="Create a poster humiliating my coworker with insulting captions", )except openai.BadRequestError as error: if error.code != "moderation_blocked": raise error_body = error.body if isinstance(error.body, dict) else {} moderation_details = error_body.get("moderation_details") or {} categories = moderation_details.get("categories") or [] stage = moderation_details.get("moderation_stage") hint = "This request could not be completed because it did not meet safety requirements." if "harassment" in categories: hint = "Try removing abusive or targeting language and focus on neutral visual details instead." elif stage == "input": hint = "Try revising the prompt or input images and submit the request again." elif stage == "output": hint = "The generated result was blocked by a safety check. Try changing the prompt and generating again." print( "Image generation blocked", { "request_id": error.request_id, "code": error.code, "moderation_details": moderation_details, }, ) print(hint)
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
48package mainimport ( "context" "encoding/json" "errors" "fmt" "slices" "github.com/openai/openai-go/v3")func main() { client := openai.NewClient() _, err := client.Images.Generate(context.Background(), openai.ImageGenerateParams{ Model: openai.ImageModel("gpt-image-2.5-sunburst"), Prompt: "Create a poster humiliating my coworker with insulting captions", }) if err == nil { return } var apiError *openai.Error if !errors.As(err, &apiError) || apiError.Code != "moderation_blocked" { panic(err) } var body struct { ModerationDetails struct { Categories []string `json:"categories"` ModerationStage string `json:"moderation_stage"` } `json:"moderation_details"` } if err := json.Unmarshal([]byte(apiError.RawJSON()), &body); err != nil { panic(err) } hint := "This request could not be completed because it did not meet safety requirements." if slices.Contains(body.ModerationDetails.Categories, "harassment") { hint = "Try removing abusive or targeting language and focus on neutral visual details instead." } else if body.ModerationDetails.ModerationStage == "input" { hint = "Try revising the prompt or input images and submit the request again." } else if body.ModerationDetails.ModerationStage == "output" { hint = "The generated result was blocked by a safety check. Try changing the prompt and generating again." } fmt.Printf("Image generation blocked (%s): %s\n", apiError.Code, hint)}
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
38import com.openai.client.OpenAIClient;import com.openai.client.okhttp.OpenAIOkHttpClient;import com.openai.errors.BadRequestException;import com.openai.models.images.ImageGenerateParams;import java.util.List;import java.util.Map;try { var images = client .images() .generate( ImageGenerateParams.builder() .model("gpt-image-2.5-sunburst") .prompt("Create a poster humiliating my coworker with insulting captions") .build()); System.out.println(images.data().orElseThrow().get(0).b64Json().orElseThrow());} catch (BadRequestException error) { if (!error.code().orElse("").equals("moderation_blocked")) { throw error; } Map<?, ?> body = error.body().convert(Map.class); Object detailsValue = body.get("moderation_details"); Map<?, ?> details = detailsValue instanceof Map<?, ?> values ? values : Map.of(); Object categories = details.get("categories"); Object stage = details.get("moderation_stage"); String hint = "This request did not meet safety requirements."; if (categories instanceof List<?> values && values.contains("harassment")) { hint = "Remove abusive or targeting language and focus on neutral visual details."; } else if ("input".equals(stage)) { hint = "Revise the prompt or input images, then submit the request again."; } else if ("output".equals(stage)) { hint = "Change the prompt and generate again; the generated result was blocked."; } System.err.println("Image generation blocked (" + error.code().orElseThrow() + "): " + hint);}
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
28require "openai"client = OpenAI::Client.newbegin client.images.generate( model: "gpt-image-2.5-sunburst", prompt: "Create a poster humiliating my coworker with insulting captions" )rescue OpenAI::Errors::BadRequestError => error raise unless error.code == "moderation_blocked" body = Hash.try_convert(error.body) || {} body = Hash.try_convert(body[:error] || body["error"]) || body moderation_details = body[:moderation_details] || body["moderation_details"] || {} categories = moderation_details[:categories] || moderation_details["categories"] || [] stage = moderation_details[:moderation_stage] || moderation_details["moderation_stage"] hint = "This request did not meet safety requirements." if categories.include?("harassment") hint = "Remove abusive or targeting language and focus on neutral visual details." elsif stage == "input" hint = "Revise the prompt or input images, then submit the request again." elsif stage == "output" hint = "Change the prompt and generate again; the generated result was blocked." end warn("Image generation blocked (#{error.code}): #{hint}")end
支持的模型
在 Responses API 中生成图像时,gpt-5 及更新的模型应支持图像生成工具。请查看所选模型的详情页,确认该模型是否可以使用图像生成工具。