文章

AI - gtp-image-2使用（openrouter）

发表于 更新于
作者 晓宇
5 分钟阅读
AI - gtp-image-2使用（openrouter）

gtp-image-2使用（openrouter）

参考文档：https://openrouter.ai/docs/client-sdks/go/api-reference/videogeneration

The Go SDK and docs are currently in beta. Report issues on GitHub.

https://openrouter.ai/models?q=gpt&output_modalities=image

因为sdk是beta测试版，干脆按照openai原协议进行对接生图。

两种方案例子：

  1. https://openrouter.ai/api/v1/chat/completions
  2. https://openrouter.ai/api/v1/responses

对话模式

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

POST https://openrouter.ai/api/v1/chat/completions
Authorization: Bearer 
Content-Type: application/json

{
  "model": "openai/gpt-5.4-image-2",
  "messages": [
    {
      "role": "user",
      "content": [
      	{
      	"type":"text",
      	"text":"a b c d e fg "
      	},
      	{
      	"type":"image_url",
      	"image_url":"https://www.360club.net/pic/124/jpg/1.jpg"
      	}
      ]
    }
  ],
  "modalities": [
    "image",
    "text"
  ],
  "image_config": {
      "quality": "high",
      "aspect_ratio": "16:9",
      "image_size": "2K"          // [可选] 控制图像分辨率，如 "1K", "2K", "4K"
  }
}

输出：

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

{
	"id": "gen-1778815970-WfDVKNLvEAn8BEWfkHIA",
	"object": "chat.completion",
	"created": 1778815970,
	"model": "openai/gpt-5.4-image-2-20260421",
	"provider": "OpenAI",
	"system_fingerprint": null,
	"service_tier": "default",
	"choices": [{
		"index": 0,
		"logprobs": null,
		"finish_reason": "stop",
		"native_finish_reason": "completed",
		"message": {
			"role": "assistant",
			"content": null,
			"refusal": null,
			"reasoning": null,
			"images": [{
				"type": "image_url",
				"image_url": {
					"url": "data:image/png;base64,"
				}
			}]
		}
	}],
	"usage": {
		"prompt_tokens": 1635,
		"completion_tokens": 7067,
		"total_tokens": 8702,
		"cost": 0.224445,
		"is_byok": false,
		"prompt_tokens_details": {
			"cached_tokens": 0,
			"cache_write_tokens": 0,
			"audio_tokens": 0,
			"video_tokens": 0
		},
		"cost_details": {
			"upstream_inference_cost": 0.224445,
			"upstream_inference_prompt_cost": 0.01308,
			"upstream_inference_completions_cost": 0.211365
		},
		"completion_tokens_details": {
			"reasoning_tokens": 0,
			"image_tokens": 7024,
			"audio_tokens": 0
		}
	}
}

续写模式

https://developers.openai.com/api/reference/resources/responses/methods/create

参考文档的一些枚举值，确保可以请求通过。

模型名映射关系：

  • openai/gpt-5.4-image-2 => openai/gpt-5.4-image-2-20260421
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

POST https://openrouter.ai/api/v1/responses
Authorization: Bearer 
Content-Type: application/json 

{
  "model" : "gpt-5.4",
  "input" : [
    {
    "role" : "user",
    "content" : [{
      "type" : "input_text",
      "text" : "Theme and Style: Create a set of high-quality 3D cute cartoon avatars (personal portraits) based on characters in reference images. The aesthetic style is :Memoji style in Apple's iOS system.\n\nCharacters and Expressions: Highly accurate representation of the core features of the characters. The final product should include 9 different versions of the avatars, each displaying a unique, vibrant pose and distinct facial expression, presenting a lively and adorable image. Some avatars will carry props that match their emotions.\n\nComposition and Layout: Use a simple 3x3 grid layout, containing 9 independent avatars. All scenes are placed on a seamless pure white background. The key is to use only white space to separate the characters; there should be no visible grid lines, borders, frames, or boxes between characters.\n\nDetails and Styling: The 9 virtual characters will be randomly paired with some small accessories.\n\nImportant Visual Constraints: Grid lines, black lines, cell borders, borders, distorted facial features, and text are prohibited.\n And  1k resolution."
    }, {
      "type" : "input_image",
      "image_url" : "https://www.360club.net/pic/124/jpg/1.jpg"
    } ]
  } ],
  "tools" : [ {
    "type" : "image_generation",
    "action" : "auto",
    "size" : "1024x1024",
    "output_format" : "png"
  } ]
}

输出：

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

{
	"id": "gen-1778938920-zXWbSt7ffeet8ywZH4Vx",
	"object": "response",
	"created_at": 1778938920,
	"model": "openai/gpt-5.4-20260305",
	"status": "completed",
	"completed_at": 1778939034,
	"output": [{
		"type": "openrouter:image_generation",
		"id": "st_tmp_5bru8cqyzcw",
		"status": "completed",
		"result": "data:image/png;base64,=="
	}, {
		"id": "msg_tmp_ehp6pv4yq9",
		"type": "message",
		"role": "assistant",
		"status": "completed",
		"content": [{
			"type": "output_text",
			"text": "Done.",
			"annotations": [],
			"logprobs": []
		}]
	}],
	"error": null,
	"incomplete_details": null,
	"tools": [{
		"type": "image_generation"
	}],
	"tool_choice": "auto",
	"parallel_tool_calls": true,
	"max_output_tokens": null,
	"temperature": 1,
	"top_p": 1,
	"presence_penalty": 0,
	"frequency_penalty": 0,
	"top_logprobs": 0,
	"max_tool_calls": null,
	"metadata": {},
	"background": false,
	"previous_response_id": null,
	"service_tier": "auto",
	"truncation": "disabled",
	"store": false,
	"instructions": null,
	"text": {
		"format": {
			"type": "text"
		}
	},
	"reasoning": null,
	"safety_identifier": null,
	"prompt_cache_key": null,
	"usage": {
		"input_tokens": 3932,
		"input_tokens_details": {
			"cached_tokens": 0
		},
		"output_tokens": 253,
		"output_tokens_details": {
			"reasoning_tokens": 0
		},
		"total_tokens": 4185,
		"cost": 0.013625,
		"cost_details": {
			"upstream_inference_cost": 0.013625,
			"upstream_inference_input_cost": 0.00983,
			"upstream_inference_output_cost": 0.0037949999999999998
		}
	}
}

结论

使用对话模式时，需要使用返回的image字段，如果使用golang的openai的库：github.com/openai/openai-go/v3时，是没有这个参数接收的（也就是内部接收数据不全，需要自己实现）– openrouter官方建议对话模式。

​

使用续写模式时，参数的model却不能跟openrouter一致，必须用openai的，这样程序上就需要兼容或者使用两套来对比效果。

ai, openrouter
ai openrouter
本文由作者按照 CC BY 4.0 进行授权