Multimodal-Embedding API详情

图片向量化示例

import dashscope
import json
from http import HTTPStatus
# 实际使用中请将url地址替换为您的图片url地址
image = "https://dashscope.oss-cn-beijing.aliyuncs.com/images/256_1.png"
input = [{'image': image}]
# 调用模型接口
resp = dashscope.MultiModalEmbedding.call(
    model="tongyi-embedding-vision-plus",
    input=input
)

if resp.status_code == HTTPStatus.OK:
    result = {
        "status_code": resp.status_code,
        "request_id": getattr(resp, "request_id", ""),
        "code": getattr(resp, "code", ""),
        "message": getattr(resp, "message", ""),
        "output": resp.output,
        "usage": resp.usage
    }
    print(json.dumps(result, ensure_ascii=False, indent=4))

import dashscope
import base64
import json
from http import HTTPStatus
# 读取图片并转换为Base64,实际使用中请将xxx.png替换为您的图片文件名或路径
image_path = "xxx.png"
with open(image_path, "rb") as image_file:
    # 读取文件并转换为Base64
    base64_image = base64.b64encode(image_file.read()).decode('utf-8')
# 设置图像格式
image_format = "png"  # 根据实际情况修改，比如jpg、bmp 等
image_data = f"data:image/{image_format};base64,{base64_image}"
# 输入数据
input = [{'image': image_data}]

# 调用模型接口
resp = dashscope.MultiModalEmbedding.call(
    model="tongyi-embedding-vision-plus",
    input=input
)
if resp.status_code == HTTPStatus.OK:
    result = {
        "status_code": resp.status_code,
        "request_id": getattr(resp, "request_id", ""),
        "code": getattr(resp, "code", ""),
        "message": getattr(resp, "message", ""),
        "output": resp.output,
        "usage": resp.usage
    }
    print(json.dumps(result, ensure_ascii=False, indent=4))

视频向量化示例

多模态向量化模型目前仅支持以URL形式输入视频文件，暂不支持直接传入本地视频。

import dashscope
import json
from http import HTTPStatus
# 实际使用中请将url地址替换为您的视频url地址
video = "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20250107/lbcemt/new+video.mp4"
input = [{'video': video}]
# 调用模型接口
resp = dashscope.MultiModalEmbedding.call(
    model="tongyi-embedding-vision-plus",
    input=input
)

if resp.status_code == HTTPStatus.OK:
    result = {
        "status_code": resp.status_code,
        "request_id": getattr(resp, "request_id", ""),
        "code": getattr(resp, "code", ""),
        "message": getattr(resp, "message", ""),
        "output": resp.output,
        "usage": resp.usage
    }
    print(json.dumps(result, ensure_ascii=False, indent=4))
    

文本向量化示例

import dashscope
import json
from http import HTTPStatus

text = "通用多模态表征模型示例"
input = [{'text': text}]
# 调用模型接口
resp = dashscope.MultiModalEmbedding.call(
    model="tongyi-embedding-vision-plus",
    input=input
)

if resp.status_code == HTTPStatus.OK:
    result = {
        "status_code": resp.status_code,
        "request_id": getattr(resp, "request_id", ""),
        "code": getattr(resp, "code", ""),
        "message": getattr(resp, "message", ""),
        "output": resp.output,
        "usage": resp.usage
    }
    print(json.dumps(result, ensure_ascii=False, indent=4))

融合向量化示例

import dashscope
import json
import os
from http import HTTPStatus

# 多模态融合向量：将文本、图片、视频融合成一个融合向量
# 适用于跨模态检索、图搜等场景
text = "这是一段测试文本，用于生成多模态融合向量"
image = "https://dashscope.oss-cn-beijing.aliyuncs.com/images/256_1.png"
video = "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20250107/lbcemt/new+video.mp4"

# 输入包含文本、图片、视频，通过 enable_fusion 参数生成融合向量
input_data = [
    {"text": text},
    {"image": image},
    {"video": video}
]

# 使用 qwen3-vl-embedding 生成融合向量
resp = dashscope.MultiModalEmbedding.call(
    # 若没有配置环境变量，请用百炼API Key将下行替换为：api_key="sk-xxx",
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    model="qwen3-vl-embedding",
    input=input_data,
    enable_fusion=True,
    # 可选参数：指定向量维度（支持 2560、2048、1536、1024、768、512、256，默认 2560）
    # parameters={"dimension": 1024}
)

print(json.dumps(resp, ensure_ascii=False, indent=4))

多图融合向量化示例

以下示例展示如何使用 qwen3-vl-embedding 模型，将多张图片与文本融合为 1 个向量。传入多个 image 条目即可实现多图融合，适用于商品多角度图片+描述文本的综合语义检索场景。

import dashscope
import json
import os
from http import HTTPStatus

# 多图+文本融合向量：将多张商品图片和描述文本融合为 1 个向量
# 适用于商品多角度图片+描述文本的综合语义检索
text = "白色运动鞋，轻量透气，适合跑步和日常穿着"
image1 = "https://dashscope.oss-cn-beijing.aliyuncs.com/images/256_1.png"
image2 = "https://img.alicdn.com/imgextra/i3/O1CN01rdstgY1uiZWt8gqSL_!!6000000006071-0-tps-1970-356.jpg"

# 传入多个 image 条目实现多图融合，enable_fusion=True 将所有输入融合为 1 个向量
input_data = [
    {"text": text},
    {"image": image1},
    {"image": image2}
]

resp = dashscope.MultiModalEmbedding.call(
    # 若没有配置环境变量，请用百炼API Key将下行替换为：api_key="sk-xxx",
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    model="qwen3-vl-embedding",
    input=input_data,
    enable_fusion=True
)

print(json.dumps(resp, ensure_ascii=False, indent=4))

2026-03-06 快照版本示例

以下示例展示如何使用 tongyi-embedding-vision-plus-2026-03-06 模型，演示 res_level（多分辨率）和 max_video_frames（视频帧数）参数的使用。该模型基于 Qwen3 底座，支持 30+ 种语言，同时支持独立向量和融合向量生成。

import dashscope
import json
import os
from http import HTTPStatus

# tongyi-embedding-vision-plus-2026-03-06 示例
# 支持 res_level（多分辨率）和 max_video_frames（视频帧数）参数
image = "https://dashscope.oss-cn-beijing.aliyuncs.com/images/256_1.png"
video = "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20250107/lbcemt/new+video.mp4"
text = "这是1个视觉多模态表征模型"

input_data = [
    {"text": text},
    {"image": image},
    {"video": video}
]

# 调用 2026-03-06 快照版本模型
resp = dashscope.MultiModalEmbedding.call(
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    model="tongyi-embedding-vision-plus-2026-03-06",
    input=input_data,
    dimension=1152,      # 支持 1152/1024/512/256/128/64
    res_level=1,         # 分辨率档位：0/1/2/3，默认 1
    max_video_frames=64  # 视频最大采样帧数，默认 8，最大 64
)

if resp.status_code == HTTPStatus.OK:
    result = {
        "status_code": resp.status_code,
        "request_id": getattr(resp, "request_id", ""),
        "output": resp.output,
        "usage": resp.usage
    }
    print(json.dumps(result, ensure_ascii=False, indent=4))

以下示例展示 2026-03-06 版本的融合向量用法：将 text、image、video 放在同一个 content 对象中，模型会将所有输入融合编码为 1 个向量（type 为 fused）。

import dashscope
import json
import os
from http import HTTPStatus

# 融合向量：将 text/image/video 放在同一个 content 对象中
# 模型会将所有输入融合编码为 1 个向量，type 为 "fused"
text = "白色运动鞋，轻量透气，适合跑步和日常穿着"
image = "https://dashscope.oss-cn-beijing.aliyuncs.com/images/256_1.png"

# 同一对象中的多模态内容会被融合为 1 个向量
input_data = [
    {"text": text, "image": image}
]

resp = dashscope.MultiModalEmbedding.call(
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    model="tongyi-embedding-vision-plus-2026-03-06",
    input=input_data,
    dimension=1152
)

if resp.status_code == HTTPStatus.OK:
    result = {
        "status_code": resp.status_code,
        "request_id": getattr(resp, "request_id", ""),
        "output": resp.output,
        "usage": resp.usage
    }
    print(json.dumps(result, ensure_ascii=False, indent=4))