Platform

Pricing

Solutions

Build

Resources

Company

Select Language

Try Free

Talk to Sales

PEGASUS 1.6

media & entertainment

Understand video from a first-person perspective.

A video-to-text model that returns time-coded JSON, now with native image understanding, support for egocentric (first-person) footage, better entity recognition, and cleaner naming.

2hrs

Max video duration with full temporal context retained end-to-end.

Languages supported across input prompts and generated output.

Pre-indexing steps. Send a URL, an asset, or base64, get text back.

JSON

Structured segmentation output ready for your editor, or pipeline.

Things only Pegasus does.

General-purpose VLMs paste frames together and call it understanding. Pegasus is built for video and images, and the gap shows as soon as your prompt mentions a timestamp.

Understands first-person footage.

Bring video descriptions and analysis into the perspective of the person behind the camera with Pegasus 1.6’s expanded understanding of egocentric point-of-view.

Reads the fine print.

On-screen text. Jersey numbers. Whiteboards. Receipts. Pegasus parses the frame alongside the speech track, so summaries include the slides too.

From a raw video to structured JSON.

Define a segment such as a speaker change, brand appearance, or scene cut, choose your fields, and Pegasus returns timestamped JSON.

Sees the frame, not just the script.

Pegasus analyzes images natively, giving it direct access to on-screen text, logos, and visual detail. No bolted-on OCR model, no latency or context-window tax.

The answer comes with a timestamp.

Pegasus answers with exact timestamp, not “somewhere in the middle.” Granular temporal reasoning is built into the model.

Knows who's who.

Improve entity recognition and naming, narrowing the gap with leading models on character/entity ID, which is critical for content ID and character recognition.

From signup to first result in 5 minutes.

Same model, prompts, and JSON output. Choose the surface for your team.

Python

Node.js

import requests
 
# Step 2: Define the API URL and the specific endpoint
API_URL = "https://api.twelvelabs.io/v1.3"
INDEXES_URL = f"{API_URL}/indexes"
 
# Step 3: Create the necessary headers for authentication
headers = {
    "x-api-key": "<YOUR_API_KEY>"
}
 
# Step 4: Prepare the data payload for your API request
INDEX_NAME = "<YOUR_INDEX_NAME>"
data = {
    "models": [
        {
            "model_name": "marengo3.0",
            "model_options": ["visual", "audio"]
        }
    ]
}

Quick Start

Index your first video and run a search in under 5 minutes.

Sample Apps

Production repos: video search, RAG over video, highlight reel generator, compliance scanner.

MCP Server

Connect Claude, Cursor, or any MCP client to your video index.

SDKs & API

Python, Node, REST. Full reference, typed responses, streaming.

Quick Start

Index your first video and run a search in under 5 minutes.

Sample Apps

Production repos: video search, RAG over video, highlight reel generator, compliance scanner.

MCP Server

Connect Claude, Cursor, or any MCP client to your video index.

SDKs & API

Python, Node, REST. Full reference, typed responses, streaming.

Built for video, now with native image support.

General-purpose multimodal LLMs handle video by sampling stills and creating captions. Pegasus processes the entire video with native image understanding that reads on-screen text, logos, and visual details directly.

CAPABILITY

PEGASUS 1.6

Gemini 3.1 PRO

GPT-5.5

Max single-call duration

120 min

90 min

Not specified (omnimodal, no published video duration cap)

Structured segmentation output

JSON-native, schema-conditioned

Structured Outputs supported; no native temporal segmentation

Multimodal prompting (image+text)