Video Input Implementation Guide - SambaNova Documentation

SambaNova supports video input for select multimodal models on both SambaCloud and SambaStack 1.2, allowing you to pass video files alongside text prompts. Learn how to query SambaNova video-capable models using either the SambaNova or OpenAI Python client.

Video input is currently supported on gemma-4-31B-it. This model does not support audio input.

Videos must meet the following requirements:

Maximum duration: 60 seconds
Frame rate: 2 fps
Maximum file size: 100 MB
Duration metadata: The video file must have readable duration metadata

Always use "type": "video_url" with a nested video_url object. Using "type": "video" is not supported and causes a validation error.

Make a query with a video

The video request follows the multimodal input format. Pass the video as a base64-encoded string or a URL using the video_url content type.

Step 1

Make a new Python file and copy the code below.

from sambanova import SambaNova
import base64

client = SambaNova(
    base_url="your-sambanova-base-url",
    api_key="your-sambanova-api-key",
)

def encode_video(video_path):
    with open(video_path, "rb") as video_file:
        return base64.b64encode(video_file.read()).decode("utf-8")

video_path = "sample.mp4"
video_base64 = encode_video(video_path)

response = client.chat.completions.create(
    model="gemma-4-31B-it",
    messages=[
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "Describe what is happening in this video."},
                {
                    "type": "video_url",
                    "video_url": {"url": f"data:video/mp4;base64,{video_base64}"},
                },
            ],
        }
    ],
)

print(response.choices[0].message.content)

from openai import OpenAI
import base64

client = OpenAI(
    base_url="your-sambanova-base-url",
    api_key="your-sambanova-api-key",
)

def encode_video(video_path):
    with open(video_path, "rb") as video_file:
        return base64.b64encode(video_file.read()).decode("utf-8")

video_path = "sample.mp4"
video_base64 = encode_video(video_path)

response = client.chat.completions.create(
    model="gemma-4-31B-it",
    messages=[
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "Describe what is happening in this video."},
                {
                    "type": "video_url",
                    "video_url": {"url": f"data:video/mp4;base64,{video_base64}"},
                },
            ],
        }
    ],
)

print(response.choices[0].message.content)

Step 2

Use your SambaNova API key and base URL from the API keys and URLs page to replace "your-sambanova-api-key" and "your-sambanova-base-url".

Step 3

Replace "sample.mp4" with the path to your video file.

Step 4

Update the prompt text in the content array to match your use case.

Step 5

Run the Python file to receive the text output.

Payload format

The video_url content block accepts either a base64-encoded video or a URL:

{
  "type": "video_url",
  "video_url": {
    "url": "data:video/mp4;base64,<base64-encoded-video>"
  }
}

{
  "type": "video_url",
  "video_url": {
    "url": "https://example.com/sample.mp4"
  }
}

Get started

Models

Features

Build

Resources

Implement video input features

Make a query with a video

Payload format

​Make a query with a video

​Payload format

Make a query with a video

Payload format