Multimodal embedding API

Image embedding

import dashscope
import json
from http import HTTPStatus
# Replace with your image URL.
image = "https://dashscope.oss-cn-beijing.aliyuncs.com/images/256_1.png"
input = [{'image': image}]
# Call the model API.
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
# Read the image and convert it to Base64. Replace xxx.png with your image file.
image_path = "xxx.png"
with open(image_path, "rb") as image_file:
    # Read the file and convert it to Base64.
    base64_image = base64.b64encode(image_file.read()).decode('utf-8')
# Set the image format.
image_format = "png"  # Change this to your image's format (e.g., jpg, bmp).
image_data = f"data:image/{image_format};base64,{base64_image}"
# Input data
input = [{'image': image_data}]

# Call the model API.
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))

Video embedding

Currently, the model only supports video input via URL. Local video files are not supported.

import dashscope
import json
from http import HTTPStatus
# Replace with your video URL.
video = "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20250107/lbcemt/new+video.mp4"
input = [{'video': video}]
# Call the model API.
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))
    

Text embedding

import dashscope
import json
from http import HTTPStatus

text = "General multimodal representation model example"
input = [{'text': text}]
# Call the model API.
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))

Fused embedding

import dashscope
import json
import os
from http import HTTPStatus

# Fuses text, image, and video into a single fused embedding.
# Ideal for use cases like cross-modal retrieval and image search.
text = "This is a test text for generating a multimodal fused embedding."
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"

# Input includes text, image, and video. Set enable_fusion=True to generate a fused embedding.
input_data = [
    {"text": text},
    {"image": image},
    {"video": video}
]

resp = dashscope.MultiModalEmbedding.call(
    # If the environment variable is not set, provide your Model Studio API key, e.g., api_key="sk-xxx".
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    model="qwen3-vl-embedding",
    input=input_data,
    enable_fusion=True,
    # Optional: Specify the embedding dimension. Valid values: 2560, 2048, 1536, 1024, 768, 512, and 256. Default: 2560.
    # parameters={"dimension": 1024}
)

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

Multi-image fused embedding

Use qwen3-vl-embedding to fuse multiple images and text into a single embedding. To fuse multiple images, pass multiple image items. This is ideal for semantic retrieval using multi-angle product images and a text description.

import dashscope
import json
import os
from http import HTTPStatus

# Fuses multiple product images and a description into a single embedding.
# Ideal for comprehensive semantic retrieval using multi-angle product images and a text description.
text = "White sports shoes, lightweight and breathable, suitable for running and daily wear."
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"

# Pass multiple image items and set enable_fusion=True to fuse all inputs into a single embedding.
input_data = [
    {"text": text},
    {"image": image1},
    {"image": image2}
]

resp = dashscope.MultiModalEmbedding.call(
    # If the environment variable is not set, provide your Model Studio API key, e.g., 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 snapshot version

This example shows how to use the tongyi-embedding-vision-plus-2026-03-06 model and its res_level (resolution) and max_video_frames (video frames) parameters. Built on the Qwen3 foundation, this model supports 30+ languages and generates both independent and fused embeddings.

import dashscope
import json
import os
from http import HTTPStatus

# Demonstrates using the res_level (resolution) and max_video_frames (video frames) parameters.
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 = "This is a visual multimodal representation model."

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

resp = dashscope.MultiModalEmbedding.call(
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    model="tongyi-embedding-vision-plus-2026-03-06",
    input=input_data,
    dimension=1152,      # Valid values: 1152, 1024, 512, 256, 128, 64
    res_level=1,         # Resolution level: 0, 1, 2, or 3. The default is 1.
    max_video_frames=64  # Maximum number of sampled video frames. Default: 8. Maximum: 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))

To generate a fused embedding with the 2026-03-06 version, place text, image, and video in the same content object. The model fuses all inputs into a single embedding with the type fused.

import dashscope
import json
import os
from http import HTTPStatus

# To create a fused embedding, place text and image in the same content object.
# The model fuses all inputs into a single embedding of type `fused`.
text = "White sports shoes, lightweight and breathable, suitable for running and daily wear."
image = "https://dashscope.oss-cn-beijing.aliyuncs.com/images/256_1.png"

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))