参数名 参数说明 是否必须 参数类型 示例 取值范围
model 模型id 是 string MAILAND/majicflus_v1 ModelScope上的AIGC 模型ID
prompt 正向提示词，大部分模型建议使用英文提示词效果较好。 是 string A mysterious girl walking down the corridor. 长度小于2000
negative_prompt 负向提示词 否 string lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry 长度小于2000
size 生成图像分辨率大小 否 string 1024x1024 分辨率范围:
SD系列:[64x64,2048x2048]，FLUX:[64x64,1024x1024]，Qwen-Image:[64x64,1664x1664]
seed 随机种子 否 int 12345 [0,2^31-1]
steps 采样步数 否 int 30 [1,100]
guidance 提示词引导系数 否 float 3.5 [1.5,20]
image_url 待编辑图片的url地址，该参数只适用于支持图片编辑的模型 否 string https://resources.modelscope.cn/aigc/image_edit.png 确保公网可访问
-----------------------------------------
import requests
import time
import json
from PIL import Image
from io import BytesIO
base_url = 'https://api-inference.modelscope.cn/'
api_key = "<MODELSCOPE_SDK_TOKEN>"
common_headers = {
"Authorization": f"Bearer {api_key}",
"Content-Type": "application/json",
}
response = requests.post(
f"{base_url}v1/images/generations",
headers={**common_headers, "X-ModelScope-Async-Mode": "true"},
data=json.dumps({
"model": "black-forest-labs/FLUX.1-Krea-dev", # ModelScope Model-Id, required
"prompt": "A golden cat"
}, ensure_ascii=False).encode('utf-8')
)
response.raise_for_status()
task_id = response.json()["task_id"]
while True:
result = requests.get(
f"{base_url}v1/tasks/{task_id}",
headers={**common_headers, "X-ModelScope-Task-Type": "image_generation"},
)
result.raise_for_status()
data = result.json()
if data["task_status"] == "SUCCEED":
image = Image.open(BytesIO(requests.get(data["output_images"][0]).content))
image.save("result_image.jpg")
break
elif data["task_status"] == "FAILED":
print("Image Generation Failed.")
break
time.sleep(5)
