A modern newsroom runs on a dozen kinds of software, usually from as many makers. A newsroom system plans the story, a graphics engine builds its lower thirds, automation runs it on air, a web CMS publishes it, and AI tools transcribe, match, and check along the way. Each kind of tool is good at its own job. The hard part is that they rarely share what they know about the story. The Story Object Model (SOM) gives them one story to share. This note goes through the categories of newsroom tools, names a few examples of each, says briefly what they are used for, and shows where each category fits in SOM 1.0.

The products named below are examples, listed in alphabetical order, to make each category concrete. They are not a ranking or a recommendation, and naming a product says nothing about whether it supports SOM.

How a tool meets SOM

Every SOM message says which kind of system sent it, in the envelope’s originating_system.system_type. SOM 1.0 has fourteen values: ncs, mos_device, graphics, automation, wire_service, ai_agent, compliance_engine, editorial_dashboard, archive, prompter, camera, audio, skill_worker, and custom. They are the closest thing SOM has to a map of the newsroom’s tools.

Two rules shape how every category fits:

One owner per story. The system that owns the story, usually the newsroom system, mints it and publishes its complete state as story.context; other tools read it.
Everyone else adds their own messages. A graphics engine commits a graphic with link.committed, automation reports that a story went to air with telling.started, a media store announces new material with delivery.media_available, an AI skill raises skill.warning.raised. The owner folds what matters into its next snapshot.

Newsroom and planning systems

What they do. The newsroom computer system (NRCS, or NCS) is where journalists write stories and scripts, and producers build and time the rundowns for each show. Planning tools keep the diary of expected events and coordinate who covers what. Many products now do both, and publish to the web and social from the same place.

Examples. AP ENPS; CGI OpenMedia; Indigo Editorial, formerly Inception (Ross Video); MediaCentral Newsroom Management, formerly iNEWS (Avid); Octopus Newsroom; Saga, formerly Dina (Fonn Group); Wolftech News (Avid).

In SOM. system_type ncs. The newsroom system is usually the story’s owner: it publishes story.context and receives everything the other tools say about the story, from skill warnings to tellings. Its rundowns are one view of the story, not the story itself.

Wire services

What they do. News agencies send a constant stream of text, photos, video, and data. Newsrooms receive it through agency platforms and APIs, and most of it never becomes a story of their own. Agencies and newsrooms describe news items with IPTC standards: NewsML-G2 in XML and ninjs in JSON.

Examples. AFP (the AFP API), AP (the AP Media API), dpa (dpa-digitalwires), and Reuters (Reuters Connect).

In SOM. A wire item a newsroom picks up is recorded in the story as an editorial_source, with source_type WIRE, the provider, a credibility level, and when it was received. SOM 1.0 has a wire_service system type, but no message family of its own for wire items: wires feed the story, and SOM does not replace their own formats.

Broadcast graphics

What they do. Graphics systems build and play out lower thirds, maps, full-frame graphics, and tickers. Journalists fill in templates from the newsroom system; operators or automation play them on air.

Examples. CasparCG Server (open source); PRIME (Chyron); Singular.live; Viz Pilot Edge and Viz Trio (Vizrt); XPression (Ross Video).

In SOM. system_type graphics. A graphic is an asset in the story (GRAPHIC, MAP, LOWER_THIRD). The graphics system reads story.context, and publishes link.committed when it commits a graphic to an output, or link.withdrawn when it pulls one. It can read the story’s editorial gates too, so a held figure can stay off screen.

Studio automation and playout

What they do. Automation turns a rundown into timed commands for the switcher, cameras, servers, graphics, and audio, so a small team can run a live show. Playout systems put channels on air.

Examples. Cinegy Air; Cuez Automator; Ignite and AMPP Automation (Grass Valley); OverDrive (Ross Video); Sofie (open source); Viz Mosart (Vizrt).

In SOM. system_type automation. Automation reads the story and the links committed to its show, and reports what actually went to air with telling.started and telling.ended. Whether something aired is read from its tellings, never stored on the asset.

Prompters

What they do. Prompters show presenters their scripts, updated live from the newsroom system as the rundown changes.

Examples. Autocue (Videndum), Autoscript WinPlus-IP News (Videndum), and CueiT (CueScript).

In SOM. system_type prompter. A prompter reads: the script is a SCRIPT asset of the story. It has nothing it needs to publish.

Studio devices: servers, camera robotics, and audio

What they do. Video servers record feeds and play clips live. Robotic heads and pedestals move studio cameras. Audio consoles mix the show.

Examples. Servers: Avid Stream IO, GV I/O (Grass Valley), Tria+ (Ross Video), XS-VIA (EVS). Camera robotics: Furio and CamBot (Ross Video), T-Motion (EVS), Vinten (Videndum). Audio: Calrec (Audiotonix), Lawo.

In SOM. SOM 1.0 has mos_device, camera, and audio system types, but these devices mostly take their cues from the rundown and automation, and nothing about that has to change. Many newsroom, graphics, automation, and prompter products work with MOS, the protocol that carries rundowns to studio devices. SOM carries the story around the newsroom; MOS keeps cueing the gallery. Our comparison of MOS and SOM explains the difference.

Media asset management and archive

What they do. A media asset management system (MAM) ingests, stores, indexes, and finds footage, lets people cut clips in a browser, and hands finished media to playout and the web. The archive keeps it for years.

Examples. Dalet Flex; Iconik (Backlight); MediaCentral (Avid); Mimir (Fonn Group); Perfect Memory; Viz One (Vizrt).

In SOM. system_type archive, or custom (SOM 1.0 has no value for a MAM). When material lands, the MAM publishes delivery.media_available. The media never travels the bus: the story points at it by reference and time range.

What they do. A digital CMS is where the web desk writes, edits, and publishes articles, and sends them to apps, social platforms, and partners.

Examples. Arc XP; Brightspot; Cue (Stibo DX); Glide Publishing Platform; WordPress VIP.

In SOM. system_type custom. The CMS reads the same story.context as the bulletin, publishes link.committed when an asset is placed on a page, and telling.started when it goes live. A correction made once in the story reaches the web as well as the air.

Verification, provenance, and standards

What they do. Verification tools help journalists check images, video, and claims. Provenance standards attach a tamper-evident record of where media came from. Standards and legal desks decide what may go out, often with tools built in-house.

Examples. AP Verify; C2PA Content Credentials; Fact Check Explorer (Google); Full Fact AI; InVID-WeVerify verification plugin.

In SOM. A check becomes part of the story: a fact check or a detection is an assertion with a review_state, an asset records its content credential in authenticity_credential, and an editorial gate holds an asset until it is approved. A standards desk tool uses system_type compliance_engine and records clearances and suppressions as system.audit.

AI transcription, indexing, and skills

What they do. AI tools turn speech into text, index video by what is in it, match loose footage to stories, and check copy against a newsroom’s rules.

Examples. AssemblyAI; Moments Lab, formerly Newsbridge; Sonix; Speechmatics; Trint; Whisper (OpenAI, open source).

In SOM. system_type ai_agent, or skill_worker for a skill executor. A transcript is a TRANSCRIPT asset whose provenance says it was authored by a model, and a match is an assertion waiting for a person’s review. A skill that spots a problem publishes skill.warning.raised with a severity of hold, flag, or inform. It never changes the story itself: the owner and the editors decide.

Editorial dashboards and audience tools

What they do. Dashboards show editors what is happening: how stories perform, what people are searching for, what is spreading on social platforms, and early signs of breaking events.

Examples. Chartbeat; Dataminr for News; Google Trends; NewsWhip; Parse.ly.

In SOM. system_type editorial_dashboard. A dashboard is mostly a reader: from the bus it can show every story’s priority, gates, warnings, and tellings in one place. An alert that leads to a story can be recorded as one of its sources.

What does not change

Each tool keeps its job. SOM does not replace the newsroom system, the graphics engine, or the CMS. It lets them share the state of each story.
SOM does not pick vendors. On a SOM bus, any tool that can send and read JSON over HTTPS can take part, directly or through an adapter.
Device control stays where it is. MOS keeps cueing studio devices from the rundown.

Questions

What software does a TV newsroom use?

Typically a newsroom computer system for stories and rundowns, a planning tool, broadcast graphics, studio automation, prompters, video servers, a media asset management system, a web CMS, and a growing set of verification, AI, and audience tools. Most of them come from different makers.

What is an NRCS?

A newsroom computer system (NRCS, or NCS) is the software where journalists write stories and scripts and producers build the rundown for each show. In SOM it is usually the story’s owner, publishing its story.context.

Does a tool have to support SOM to work with it?

Not natively. A tool without SOM support can take part through an adapter that turns its events into SOM messages and back. Someone has to build and own that adapter: the newsroom, the maker, or a third party.

What is the system_type in SOM?

A field in every SOM message’s envelope that says which kind of system sent it. SOM 1.0 has fourteen values, from ncs and graphics to ai_agent and custom.

Product and company names are trademarks of their owners, used only to identify the products as examples of a category. Naming a product is not an endorsement and does not mean it supports SOM. Lists are alphabetical and not complete; descriptions are our own summaries of each maker’s public pages, checked in September 2026.