TAMS can find any second. Which one does your story need?

A politician is taking questions at a press conference. The feed has been recording for forty minutes, and somewhere in it is the answer your story turns on.

Edit while capture (i.e. cutting a recording before it has finished) has been part of newsroom life for years. It works because the editing tools and the storage were built to work together. Step outside that system to a web team, another site, a partner, or an AI service, and the recording becomes a file again: found, copied, and moved before anyone can use it. The Time-Addressable Media Store, or TAMS, takes edit while capture beyond a single system. It’s an open specification from BBC R&D that stores media as small, time-stamped segments in ordinary object storage behind an open API. Compatible tools that speak the API and have permission can request available time ranges from a feed while recording continues.

News is where TAMS has moved fastest. Reuters and AWS, working with the BBC, CNN, and Sky, built a TAMS-based exchange that shares live feeds between news organizations in near-real time. It won Best Innovative Use of Cloud at the 2025 Broadcast Tech Innovation Awards, and TAMS on AWS took the IABM Industry Partnership Award at IBC2025. At IBC2026, at least 20 exhibitors showed TAMS demos, including Avid.

Reaching media is getting easier, although permissions, formats, latency, and integration still take work. The harder question is which seconds the story needs, and where that knowledge lives.

What TAMS gets right

TAMS is good engineering, and the credit belongs to the BBC and the community around it. Three ideas make it work.

• Media is addressed by time. A flow ID and a time range identify the same frames, so tools can exchange references and avoid unnecessary whole-file copies.
• Media doesn't change once it's written. A retained flow and time range identify the same media, subject to access and retention policy. A new version becomes a new flow, and the sections nobody touched reuse the original media.
• Storage is ordinary. Segments sit in object storage, in the cloud, or on premises, and the store keeps track of where they are.

The project is also moving to open governance, with a technical steering committee drawn from across the industry. The BBC chose a software-style process over a formal standards body so the specification can keep pace with what people build on it. Version 8.2 arrived in August 2026.

What TAMS leaves out, on purpose

TAMS is a store, and it behaves like one. It keeps only what it needs to store and find media. AWS's reference architecture for fast-turnaround production draws the same line: the store holds the media, and an asset management layer holds the editorial context.

Newsrooms already create this kind of data: transcripts, logs, markers, who’s on screen. It lives in production systems such as MediaCentral and Content Core, and too often it stays there. BBC R&D wants to make it open. In September 2026 it published the Time-Addressable Data Store, or TADS, an early specification for the data that describes media over time: logs, transcripts, shot changes, who appears on screen. Each track of data shares a timeline with media in TAMS, a MAM, or an archive. The BBC’s reasoning will sound familiar to any newsroom engineer. Today the same content gets transcribed again for every team that needs it, and its history breaks every time an edited program is rendered.

TADS is a welcome step, but it’s still in alpha. Even when it matures, it describes what happens inside a recording. A newsroom also needs to know what the recording is for. Which story is it serving? Has it been cleared? Is it in tonight’s rundown? Has it gone out on social, and in what cut? Those answers live in the plan, the script, the rundown, and the publishing record, and they tie many recordings to a single story.

That means three kinds of system must work together. TAMS holds the media, asset management holds the data that describes it, and the systems that plan and produce the news hold the story. Without a platform that connects them, a newsroom adopting TAMS has to build and maintain those integrations itself, for as long as the system runs. The BBC has asked vendors publicly what it would take to complete a production-ready TAMS workflow. We think connecting these systems is a large part of the answer.

Where the story lives

In a newsroom, context lives with the story: the plan, the assignment, the script, the rundown, the rights, the edit, and every place it's published. That's why Avid for News is built around the story, and why Avid Content Core connects the story to the media.

Here's the workflow we're building toward, with TAMS underneath.

The planning desk creates a story for the politician's press conference and assigns a producer and an editor. Avid | Stream IO records the feed, keeps the high-resolution media on Avid NEXIS and writes media flows to TAMS, and the recording is attached to the story as it grows.

The producer doesn’t scrub through forty minutes of timeline. She searches Content Core for the politician’s answer on the policy and lands on the right moment, already attached to the story. The editor can start cutting before the recording ends. The package goes into the broadcast rundown in iNEWS, and the same material reaches the web and social teams. In Content Core, the production graph records what was used, by whom, and where it was published.

Tools can exchange references and avoid unnecessary whole-file copies; playback, processing and delivery still read or transfer the required media. The story context travels with the media the whole way.

When agents do the work

This matters more as AI agents take on production work. By an agent, we mean software that works out its own steps toward a goal. Give it a brief, and it decides what to search, what to check, and what to gather, and then does it. We cover this in our post-IBC2026 webinar, available on demand here.

Agents and TAMS suit each other. An agent can request any time range from any feed in an instant. The hard part is choosing. Asked to find the politician's answer, an agent needs to know which press conference, which story, which version has been cleared, and what has already been used. Without the story, it's guessing.

That's why we put connection first and agents second. Avid Assist, which we introduced at IBC2026, is the orchestration and governance layer for AI agents across the Avid platform, and it’s coming to Avid products. Its agents are designed to work inside the production context Content Core holds, and Avid Assist is designed to apply user permissions, approval controls, and usage visibility to supported agent workflows.

How we’re working toward it

TAMS makes production more efficient and opens up new possibilities: fewer whole-file copies, and media that the tools, teams and sites you authorize can reach while it’s still being recorded. For years we’ve made fast-turnaround news work inside a single system. TAMS gives us an open, shared way to take that further. We’re bringing it into the newsroom you already run, at the pace that suits you.

It starts at ingest. We’re connecting Avid | Stream IO to Content Core through TAMS, so live recordings can land in a time-addressable store and reach the story.

We’re bringing time-addressable media into supported Content Core workflows, starting with growing media flows in TAMS.

Content Core connects to MediaCentral today, so it sits alongside the systems you already run. Content Core’s native asset management is designed to handle media on its own, with no MediaCentral connection required. Either way, the aim is for TAMS-stored media to sit in the same library as everything else and carry the same story context.

Content Core Site Engine is where much of this comes together. It’s the part of Content Core designed to run Avid media services where the media already lives, and it’s where our TAMS work starts, by creating media flows in TAMS. Over time, the aim is for services such as rendering, mixdown, transcode, and conform to run in place too, avoiding unnecessary whole-file copies. Time-addressable storage and in-place processing point the same way: work by reference and move media only when the work needs it.

None of this asks you to give up what works. TAMS goes where it earns its place: media that needs to reach more tools, more teams, and more sites. It sits alongside Avid NEXIS and the storage you already have in the cloud, on premises, or both. Linear and digital teams work from the same recording.

Where this goes next

TAMS is a real step forward for broadcast infrastructure, and as of this article, it's still early. The specification is moving quickly, governance is settling, and what a production-ready TAMS newsroom looks like is an open question.

We think the answer starts with the story. If you're testing TAMS, building on it, or deciding whether to, we'd like to compare notes. Talk to our Engineering team.