---
**Daily Launch** · [https://dailylaunch.news](https://dailylaunch.news) · [RSS](https://dailylaunch.news/feed.xml)
---

# Pegasus 1.5 from TwelveLabs is now available in Amazon Bedrock
**AI Products** · Oct 20, 2026 · 3 min read
Source: AWS ML Blog — https://aws.amazon.com/blogs/machine-learning/pegasus-1-5-from-twelvelabs-is-now-available-in-amazon-bedrock/
### The Gist

TwelveLabs just landed Pegasus 1.5 on Amazon Bedrock, letting you turn up to two hours of video into structured JSON metadata. You can now use images to query video scenes, making massive video libraries finally programmable.

### Why It Matters

This moves video from a watch-only medium to a searchable data source. Builders can now build video-native search engines without the massive infra headache, while investors see AWS doubling down on specialized vertical models.

### Market Impact

This accelerates the Video-as-Data market, forcing generic model providers to compete with high-precision, vertical specialists like TwelveLabs.

- Build automated indexing tools for massive media archives like sports or news where manual tagging is impossible
- Create visual search apps that let users find specific moments in movies using a reference image instead of a text prompt
- Integrate video-to-JSON workflows into existing Digital Asset Management systems to eliminate manual metadata entry- Processing two-hour clips on-the-fly could lead to massive latency and cost spikes that break your unit economics
- Heavy reliance on Bedrock means you are effectively locked into the AWS ecosystem for your core video intelligence layer### ELI5

Imagine you have a mountain of video clips and you need to find every time a cat wearing a hat appears. Instead of watching hours of footage, you just ask this tool and it gives you a list of the exact timestamps for every single cat.

### Deep Dive

{"sections":[{"heading":"The Death of Manual Tagging","body":"Video metadata is traditionally a nightmare of human labor. By turning video into structured JSON via Bedrock, developers can finally treat video like a searchable database rather than a black box of pixels."},{"heading":"Multimodal is the Real Moat","body":"The ability to use a reference image to query a video scene is the real killer feature here. It is one thing to search for a blue car, but it is a massive step up to show the model a specific image and say find this."},{"heading":"AWS is Curating, Not Just Hosting","body":"This is not just about adding models. AWS is building a walled garden of specialized experts, signaling they want to be the default operating system for enterprise-grade AI applications."},{"heading":"What to Watch","body":"Keep an eye on the cost per minute of video processed. If pricing stays high, this remains a niche tool for high-value media, but if it scales, it becomes the backbone of every video platform."}]}

### Key Takeaways

- **Structured Video Data is Here** Move beyond text summaries to timestamped JSON that actually works in a production codebase.
- **The AWS Aggregator Play** AWS is cementing its role as the premier destination for specialized, vertical AI models.
- **The Latency Trap** Real-time video processing at this scale is expensive and slow, so watch the unit economics closely.


[View on website](https://dailylaunch.news/articles/pegasus-1-5-from-twelvelabs-is-now-available-in-amazon-bedro)