Video Generation

ZenMux supports calling video generation models via the Vertex AI protocol. This guide explains how to generate videos with ZenMux.

💡 About Video Generation

Video generation models can automatically produce high-quality video content from text descriptions. ZenMux aggregates leading video generation models such as Google Veo and ByteDance Seedance, allowing you to call them easily through a unified API interface.

Supported Models

The currently supported video generation models include (continuously updated):

• google/veo-3.1-generate-001 — Google Veo 3.1, supports high-quality video generation
• volcengine/doubao-seedance-1.5-pro — ByteDance Seedance 1.5 Pro
• volcengine/doubao-seedance-2 — ByteDance Seedance 2, the latest-generation video generation model

📚 More Models

Visit the ZenMux model list to search and view all available video generation models.

Text-to-Video

Generate videos directly from a text prompt using an asynchronous workflow: submit a generation request first, then poll until generation completes, and finally retrieve the result.

Google Veo 3.1

from google import genai
from google.genai import types
import time

client = genai.Client(
    api_key="$ZENMUX_API_KEY",  # Replace with your API Key
    vertexai=True,
    http_options=types.HttpOptions(
        api_version="v1",
        base_url="https://zenmux.ai/api/vertex-ai"
    )
)

# Step 1: Submit a video generation request
operation = client.models.generate_videos(
    model="google/veo-3.1-generate-001",  
    prompt="A golden retriever running on the beach at sunset"
)

# Step 2: Poll until generation is complete
while not operation.done:
    time.sleep(15)
    operation = client.operations.get(operation)

# Step 3: Retrieve the generated results
for video in operation.response.generated_videos:
    print(video)

Seedance 1.5 Pro

from google import genai
from google.genai import types
import time

client = genai.Client(
    api_key="$ZENMUX_API_KEY",  # Replace with your API Key
    vertexai=True,
    http_options=types.HttpOptions(
        api_version="v1",
        base_url="https://zenmux.ai/api/vertex-ai"
    )
)

# Step 1: Submit a video generation request
operation = client.models.generate_videos(
    model="volcengine/doubao-seedance-1.5-pro",  
    prompt="A golden retriever running on the beach at sunset"
)

# Step 2: Poll until generation is complete
while not operation.done:
    time.sleep(15)
    operation = client.operations.get(operation)

# Step 3: Retrieve the generated results
for video in operation.response.generated_videos:
    print(video)

Seedance 2

from google import genai
from google.genai import types
import time

client = genai.Client(
    api_key="$ZENMUX_API_KEY",  # Replace with your API Key
    vertexai=True,
    http_options=types.HttpOptions(
        api_version="v1",
        base_url="https://zenmux.ai/api/vertex-ai"
    )
)

# Step 1: Submit a video generation request
operation = client.models.generate_videos(
    model="volcengine/doubao-seedance-2",  
    prompt="A golden retriever running on the beach at sunset"
)

# Step 2: Poll until generation is complete
while not operation.done:
    time.sleep(15)
    operation = client.operations.get(operation)

# Step 3: Retrieve the generated results
for video in operation.response.generated_videos:
    print(video)

Image-to-Video

In addition to text-to-video, ZenMux also supports passing an image as the starting frame and generating a video using a text prompt. Provide the image data via the image parameter.

Google Veo 3.1

from google import genai
from google.genai import types
import time

client = genai.Client(
    api_key="$ZENMUX_API_KEY",  # Replace with your API Key
    vertexai=True,
    http_options=types.HttpOptions(
        api_version="v1",
        base_url="https://zenmux.ai/api/vertex-ai"
    )
)

# Read a local image
with open("input_image.png", "rb") as f:
    image_bytes = f.read()

# Step 1: Submit an image-to-video request
operation = client.models.generate_videos(
    model="google/veo-3.1-generate-001",  
    image=types.Image(image_bytes=image_bytes, mime_type="image/png"),  
    prompt="The dog stands up and runs toward the ocean waves"
)

# Step 2: Poll until generation is complete
while not operation.done:
    time.sleep(15)
    operation = client.operations.get(operation)

# Step 3: Retrieve the generated results
for video in operation.response.generated_videos:
    print(video)

Seedance 1.5 Pro

from google import genai
from google.genai import types
import time

client = genai.Client(
    api_key="$ZENMUX_API_KEY",  # Replace with your API Key
    vertexai=True,
    http_options=types.HttpOptions(
        api_version="v1",
        base_url="https://zenmux.ai/api/vertex-ai"
    )
)

# Read a local image
with open("input_image.png", "rb") as f:
    image_bytes = f.read()

# Step 1: Submit an image-to-video request
operation = client.models.generate_videos(
    model="volcengine/doubao-seedance-1.5-pro",  
    image=types.Image(image_bytes=image_bytes, mime_type="image/png"),  
    prompt="The dog stands up and runs toward the ocean waves"
)

# Step 2: Poll until generation is complete
while not operation.done:
    time.sleep(15)
    operation = client.operations.get(operation)

# Step 3: Retrieve the generated results
for video in operation.response.generated_videos:
    print(video)

Seedance 2

from google import genai
from google.genai import types
import time

client = genai.Client(
    api_key="$ZENMUX_API_KEY",  # Replace with your API Key
    vertexai=True,
    http_options=types.HttpOptions(
        api_version="v1",
        base_url="https://zenmux.ai/api/vertex-ai"
    )
)

# Read a local image
with open("input_image.png", "rb") as f:
    image_bytes = f.read()

# Step 1: Submit an image-to-video request
operation = client.models.generate_videos(
    model="volcengine/doubao-seedance-2",  
    image=types.Image(image_bytes=image_bytes, mime_type="image/png"),  
    prompt="The dog stands up and runs toward the ocean waves"
)

# Step 2: Poll until generation is complete
while not operation.done:
    time.sleep(15)
    operation = client.operations.get(operation)

# Step 3: Retrieve the generated results
for video in operation.response.generated_videos:
    print(video)

💡 Image-to-Video Notes

• The image parameter is passed via types.Image, which supports image_bytes (binary data) and mime_type (e.g., image/png, image/jpeg).
• The prompt parameter is optional and is used to describe how the content in the image should move or change.
• The image will be used as the starting frame of the video, and the model will generate subsequent animation based on the image content and the prompt.

Configuration

Required Parameters

• api_key: Your ZenMux API key
• vertexai: Must be set to true to enable the Vertex AI protocol
• base_url: ZenMux Vertex AI endpoint https://zenmux.ai/api/vertex-ai
• model: The video generation model name, such as google/veo-3.1-generate-001
• prompt: A text prompt describing the video content to generate

Optional Parameters

You can customize video properties such as aspect ratio, duration, and audio via the config parameter:

Parameter	Type	Description	Example Values
aspectRatio	str	Video aspect ratio	"16:9", "9:16", "1:1"
resolution	str	Video resolution	"720p", "1080p"
durationSeconds	int	Video duration (seconds)	5, 8, 10
generateAudio	bool	Whether to generate audio	True, False

Configuration example:

operation = client.models.generate_videos(
    model="google/veo-3.1-generate-001",
    prompt="A cat playing piano in a cozy room with warm lighting",
    config=types.GenerateVideosConfig(
        aspectRatio="16:9",       # Landscape 16:9  #
        resolution="720p",        # 720p resolution  #
        durationSeconds=8,        # 8-second video duration  #
        generateAudio=True,       # Generate a video with audio  #
    )
)

💡 Parameter Support Notes

Support for optional parameters may vary by model. If you pass a parameter value that the model does not support, the API will return an error. We recommend testing with default parameters first, then adjusting gradually.

Call Flow

Video generation is an asynchronous process with three steps:

1. Submit request (generate_videos): Send a video generation request and receive an operation object
2. Poll status (operations.get): Check generation status periodically; a 15-second interval is recommended
3. Retrieve results: When operation.done is True, get the videos from operation.response.generated_videos

⚠️ Generation Time

Video generation typically takes a while (from tens of seconds to several minutes). Please be patient while polling completes, and avoid setting the polling interval too short.

Best Practices

1. Prompt optimization: Use clear, specific scene descriptions, including subject, actions, environment, lighting, and other key elements
2. Polling interval: Use a 15-second polling interval to avoid overly frequent requests
3. Error handling: Add exception handling and a timeout mechanism to prevent infinite polling
4. Model selection: Choose the right model for your needs — Veo 3.1 excels at high-quality general video generation, while the Seedance family has unique strengths in specific scenarios