GPT Image 2 透明背景 API：生成、保存与检查 PNG/WebP

python -m pip install -U openai pillow

import base64
from pathlib import Path
from openai import OpenAI

OUTPUT_FORMAT = "png"  # 改为 "webp" 可请求 WebP
if OUTPUT_FORMAT not in {"png", "webp"}:
    raise ValueError("透明素材请选择 png 或 webp")

options = {
    "model": "gpt-image-2",
    "prompt": (
        "生成一个独立的橙色登山背包图标，保留拉链与背带细节。"
        "背景完全透明，主体完整，边缘干净。"
        "不要场景、纯色底、棋盘格、边框或投影。"
    ),
    "background": "transparent",
    "output_format": OUTPUT_FORMAT,
    "size": "1024x1024",
    "quality": "medium",
}
if OUTPUT_FORMAT == "webp":
    options["output_compression"] = 80  # 示例值，按素材调整

result = OpenAI().images.generate(**options)
if not result.data or not result.data[0].b64_json:
    raise RuntimeError("响应没有可保存的 base64 图片数据")

image_bytes = base64.b64decode(result.data[0].b64_json, validate=True)
target = Path(f"backpack.{OUTPUT_FORMAT}")
target.write_bytes(image_bytes)
print(f"已保存 {target}，共 {len(image_bytes)} 字节")

import base64
from pathlib import Path
from openai import OpenAI

client = OpenAI()
with open("product.png", "rb") as source:
    result = client.images.edit(
        model="gpt-image-2",
        image=source,
        prompt=(
            "将商品独立放在完全透明的背景上。"
            "保留商品几何形状、颜色、标签文字与比例。"
            "去除拍摄背景和地面投影，不添加棋盘格或实色背景。"
            "主体轮廓清晰，边缘不要白边或杂色。"
        ),
        background="transparent",
        output_format="png",
    )

if not result.data or not result.data[0].b64_json:
    raise RuntimeError("编辑响应没有图片数据")
Path("product-cutout.png").write_bytes(
    base64.b64decode(result.data[0].b64_json, validate=True)
)

import sys
from pathlib import Path
from PIL import Image

path = Path(sys.argv[1])
expected = {".png": "PNG", ".webp": "WEBP"}.get(path.suffix.lower())
with Image.open(path) as source:
    actual_format = source.format
    source.load()  # 确认图片能够完整解码
    rgba = source.convert("RGBA")

alpha = rgba.getchannel("A")
hist = alpha.histogram()
minimum, maximum = alpha.getextrema()
print({
    "format": actual_format,
    "size": rgba.size,
    "bytes": path.stat().st_size,
    "alpha_range": [minimum, maximum],
    "fully_transparent": hist[0],
    "partly_transparent": sum(hist[1:255]),
    "fully_opaque": hist[255],
})

if expected is None or actual_format != expected:
    raise SystemExit("格式不符：请核对请求、文件扩展名和转码步骤")
if maximum == 0:
    raise SystemExit("整张图完全透明：没有可见主体")
if minimum == 255:
    raise SystemExit("整张图完全不透明：没有透明或半透明像素")

print("存在透明度和可见像素；仍需检查主体轮廓")
for name, color in {
    "white": (255, 255, 255, 255),
    "dark": (24, 24, 24, 255),
    "blue": (20, 90, 210, 255),
}.items():
    background = Image.new("RGBA", rgba.size, color)
    preview = Image.alpha_composite(background, rgba).convert("RGB")
    preview.save(path.with_name(f"{path.stem}-preview-{name}.png"))

from PIL import Image

with Image.open("backpack.png") as source:
    source.convert("RGBA").save(
        "backpack-delivery.webp", format="WEBP", lossless=True
    )