Source: https://ailoot.ai/docs/zh-hant-tw/api/images/ Canonical: https://ailoot.ai/docs/zh-hant-tw/api/images/ Language: zh-hant-tw [API](https://ailoot.ai/docs/zh-hant-tw/) # 影像 [Markdown](https://ailoot.ai/docs/zh-hant-tw/api/images/index.md) 以 OpenAI Images 格式生成與編輯影像。 POST`https://api.ailoot.ai/v1/images/generations` 影像 ```bash curl https://api.ailoot.ai/v1/images/generations \ -H "Authorization: Bearer $AILOOT_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "gpt-image-2.5", "prompt": "水彩畫：岩石上的燈塔，黎明", "size": "1024x1024" }' ``` ```python img = client.images.generate( model="gpt-image-2.5", prompt="水彩畫：岩石上的燈塔，黎明", size="1024x1024", ) print(img.data[0].url or "base64:", (img.data[0].b64_json or "")[:40]) ``` 200 OK ```json { "created": 1790812800, "data": [ { "url": "https://…/content" } ], "usage": { "input_tokens": 18, "output_tokens": 1056 } } ``` ## 身分驗證 | 參數 | 類型 | 必填 | 說明 | | --- | --- | --- | --- | | `Authorization` | header | 必填 | 金鑰格式為`Bearer sk-…`。也可使用標頭`x-api-key`或`x-goog-api-key`。 | ## 請求本文 | 參數 | 類型 | 必填 | 說明 | | --- | --- | --- | --- | | `model` | string | 必填 | 影像模型：`gpt-image-2.5-sunburst`，`gemini-3.1-flash-image`，`grok-imagine-image-2.0`，`qwen-image-3.0-pro`，`image-01`，`gpt-image-2`，`gemini-3.1-flash-lite-image`，`grok-imagine-image`，`wan2.7-image`，`gpt-image-2.5-sunburst-2026-09-08`，`gemini-3-pro-image`，`grok-imagine-image-quality`另有[4](https://ailoot.ai/models/zh-hant-tw/?format=images)。 | | `prompt` | string | 必填 | 影像描述。 | | `n` | integer | 選填 | 生成影像的數量。每張完成的影像都會計費。 | | `size` | string | 選填 | 尺寸，例如`1024x1024`。允許的尺寸取決於模型。 | | `quality` | string | 選填 | 品質設定，僅限模型支援時 (例如`low`，`medium`，`high`). | | `response_format` | string | 選填 | `url`或`b64_json` —僅限模型支援選擇格式時。 | ## 編輯 POST`https://api.ailoot.ai/v1/images/edits` 請求`multipart/form-data`：欄位`image` (單張或多張影像), `prompt`，`model`，必要時加上`mask`。 ## 回應 | 參數 | 類型 | 必填 | 說明 | | --- | --- | --- | --- | | `data[]` | array | | 完成的影像：連結`url`或資料`b64_json`，依模型而定。 | 生成前，Ailoot 會保留成果的完整費用。影像費用比文字高，請在餘額中保留足夠餘裕。每張影像或影像 token 的價格，請見[目錄](https://ailoot.ai/models/zh-hant-tw/?kind=image)。 ## 編輯 · multipart/form-data ```bash : "${AILOOT_IMAGE_EDIT_MODEL:?Select a model whose actual card or GET /v1/models method supports /v1/images/edits}" curl --fail-with-body --max-time 300 \ -D image-edit-headers.txt -o image-edit.json \ https://api.ailoot.ai/v1/images/edits \ -H "Authorization: Bearer $AILOOT_API_KEY" \ -F "model=$AILOOT_IMAGE_EDIT_MODEL" \ -F 'prompt=Keep the subject and replace the background with a sunset' \ -F 'image=@input.png' ``` `input.png` → `image-edit.json` → `data[]` → `url | b64_json` ```python import base64 import json import mimetypes import os from pathlib import Path import time from urllib.parse import urljoin, urlsplit from urllib.request import Request, urlopen origin = "https://api.ailoot.ai" out = Path("image-edit-results") out.mkdir(exist_ok=True) def save_all(data): rows = data.get("data", []) if not rows: raise RuntimeError("No completed image data; keep the original response") for index, row in enumerate(rows, 1): if row.get("b64_json"): content = base64.b64decode(row["b64_json"], validate=True) suffix = ".png" if content.startswith(b"\x89PNG") else ".bin" elif row.get("url"): url = row["url"] if urlsplit(url).scheme != "https": raise ValueError("Expected the exact returned HTTPS image URL") # Signed result URLs are downloaded without sending the API key to another host. with urlopen(url, timeout=300) as response: content = response.read() suffix = mimetypes.guess_extension(response.headers.get_content_type()) or ".bin" else: raise RuntimeError("Image has neither url nor b64_json") (out / ("image-%03d" % index + suffix)).write_bytes(content) with open("image-edit.json", encoding="utf-8") as response: save_all(json.load(response)) ``` ## 影像 · async OpenAI與Anthropic SDK 通常會自動重試網路錯誤與5xx 錯誤。生成內容時，請停用此行為 (`max_retries=0`)，並在再次嘗試前檢查`X-Should-Retry`。回應遺失不代表請求未執行。 ```python import base64 import json import mimetypes import os from pathlib import Path import time from urllib.parse import urljoin, urlsplit from urllib.request import Request, urlopen origin = "https://api.ailoot.ai" key = os.environ["AILOOT_API_KEY"] out = Path("image-results") out.mkdir(exist_ok=True) def api(url, payload=None): url = urljoin(origin, url) if urlsplit(url).scheme != "https" or urlsplit(url).netloc != "api.ailoot.ai": raise ValueError("Unexpected API origin") request = Request(url, data=None if payload is None else json.dumps(payload).encode(), headers={"Authorization": "Bearer " + key, "Content-Type": "application/json"}) with urlopen(request, timeout=300) as response: headers = dict(response.headers) data = json.load(response) (out / "response-headers.json").write_text(json.dumps(headers, indent=2)) return data, headers def save_all(data): rows = data.get("data", []) if not rows: raise RuntimeError("No completed image data; keep the original response") for index, row in enumerate(rows, 1): if row.get("b64_json"): content = base64.b64decode(row["b64_json"], validate=True) suffix = ".png" if content.startswith(b"\x89PNG") else ".bin" elif row.get("url"): url = row["url"] if urlsplit(url).scheme != "https": raise ValueError("Expected the exact returned HTTPS image URL") # Signed result URLs are downloaded without sending the API key to another host. with urlopen(url, timeout=300) as response: content = response.read() suffix = mimetypes.guess_extension(response.headers.get_content_type()) or ".bin" else: raise RuntimeError("Image has neither url nor b64_json") (out / ("image-%03d" % index + suffix)).write_bytes(content) # One POST. If its result is unknown, preserve it and check history; do not rerun creation. task, headers = api("/v1/images/generations/async", { "model": "gpt-image-2.5", "prompt": "Watercolor lighthouse at dawn", "size": "1024x1024"}) (out / "created-task.json").write_text(json.dumps(task, indent=2)) task_id = task.get("task_id") or task.get("id") poll_url = task.get("poll_url") or headers.get("Location") or headers.get("location") if not task_id or not poll_url: raise RuntimeError("Keep the create response and Request ID; do not submit again") for poll in range(60): task, headers = api(poll_url) # Read the original task with the same key. (out / "task-status.json").write_text(json.dumps(task, indent=2)) if task.get("status") == "completed": result = task["result"] (out / "image-response.json").write_text(json.dumps(result, indent=2)) save_all(result) break if task.get("status") == "failed": raise RuntimeError(json.dumps(task.get("error"))) if task.get("status") != "processing" or task.get("result_pending_check"): raise RuntimeError("Keep the original task ID and check it later; no new POST") time.sleep(min(60, max(1, int(headers.get("Retry-After", headers.get("retry-after", "3")))))) else: raise TimeoutError("Polling stopped; resume GET reads using created-task.json, never repeat POST") ``` 還有疑問嗎？請提供 Request ID與問題描述，我們會協助釐清。[聯絡支援](mailto:support@ailoot.ai) ## Ailoot · 公司及其專案組合 Ailoot以累積的營運經驗、廣大的客戶群、穩定的服務及持續的KPI管理為基礎。 公司及其專案組合：數百萬客戶，每月銷售額達數十億。 Ailoot將這些營運經驗用於API存取，讓客戶的產品、代理、應用程式及自動化流程使用AI功能。