TwelveLabs launches Pegasus 1.5 for time-based metadata

// 45d agoMODEL RELEASE

TwelveLabs launches Pegasus 1.5 for time-based metadata

Pegasus 1.5 is TwelveLabs’ new video model for converting raw footage into structured, timestamped metadata rather than just answering questions about clips. The release centers on a schema-first `/analyze` workflow where teams define what matters in their domain, then get back non-overlapping temporal segments with JSON outputs that can feed search, analytics, compliance, and automation pipelines. TwelveLabs positions it as a shift from clip-based QA to production-ready video data.

// ANALYSIS

Hot take: this is a real product shift, not a cosmetic model bump. TwelveLabs is trying to make video behave more like a database table than a blob of media.

–The core change is boundary detection plus structured extraction across an entire video, which is much more operationally useful than clip-level Q&A.
–The schema-first design is the right move for enterprise workflows where “what matters” differs by domain, whether that is sports, media, or brand monitoring.
–The model is clearly aimed at downstream systems, not demos: the emphasis on valid JSON, timestamps, and non-overlapping segments matters more than fluent prose.
–The main moat is not just model quality, but the evaluation and training stack built around temporal metrics and verifiable rewards.
–If Pegasus 1.5 holds up in practice, it could reduce a lot of custom preprocessing and annotation glue code for teams working with large video libraries.

// TAGS

videomultimodalmetadatafoundation modelvideo-aienterprise-aitemporal-segmentation

DISCOVERED

45d ago

2026-04-20

PUBLISHED

45d ago

2026-04-20

RELEVANCE

9/ 10

AUTHOR

[REDACTED]

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

LAUNCH17m ago

ElectricSQL co-founder Sam Willis introduces Electric Agents, advocating for a state-sync architecture that moves AI agents out of isolated sandboxes and into durable, stream-based functions and logs.

ElectricSQL co-founder Sam Willis announced the architectural design of Electric Agents, a durable runtime platform built on a state-sync fabric. Moving away from the transient nature of traditional sandboxed execution, this architecture splits agent state and compute by utilizing durable event logs (Durable Streams) and typed, addressable entities. Under this model, every action, tool call, and state transition is appended to a persistent stream, enabling features such as session replaying, low-overhead stream forking, and the ability for agents to safely scale to zero and resume seamlessly.

LAUNCH23m ago

Electric SQL launches Electric Agents, challenging the VM sandbox status quo with a stream-backed, serverless architecture for durable agent workflows.

Electric SQL has launched Electric Agents, a platform designed for building durable, serverless AI agents. Instead of running agentic workflows inside expensive, stateful VM sandboxes, Electric Agents introduces a stateless model where agent logic runs inside lightweight serverless functions. Each agent is modeled as an addressable entity backed by an "Electric Stream"—a persistent, append-only event log acting as the agent's memory, inbox, and audit trail. This enables agents to scale to zero, sleep when idle, survive restarts, and run continuously in production while keeping compute stateless and cost-efficient.

LAUNCH31m ago

Higgsfield AI launches official Figma plugin

Higgsfield AI has released an official Figma plugin that integrates comprehensive AI-driven image and vector creation directly within the Figma design workspace. The tool allows designers to generate clean SVGs, remove backgrounds, apply color grades, swap faces while keeping scenes consistent, shoot studio product photos, and animate hero creatives, thereby eliminating the need to context-switch between Figma and external creative apps.