Transcoding moves to the cloud. Intercom, tally and shading stay at the venue

For a broadcast engineering director drawing up next year's facility plan, the useful question is no longer how much to move to the cloud. It's which workloads move, and what holds the two halves together. Dak Dillon makes a related case in a September 30 NewscastStudio analysis. His piece uses responses collected for a recent Industry Insights roundtable that looked at where live production stands today. Vendors taking part covered production, monitoring, transport, cloud and infrastructure. Dillon's view is that the industry has mostly stopped arguing about whether to distribute live production. What's still open is where distribution actually works.

Where vendors draw the cloud line

Ask one vendor after another and you get the same boundary. Transcoding and graphics are software jobs whose demand rises and falls, and they move easily. Intercom and camera shading are tied to latency, to physical presence or to both, and they don't.

• Moves relatively easily: transcoding, packaging, graphics, overlays, monitoring
• Resists the move: camera shading, audio mixing, intercom, IFB, tally

Ian Wagdin, vice president of technology and innovation at Appear, explains the split in infrastructure terms. He doesn't see hardware and virtualization as rival end states. Instead, organizations increasingly weigh each workload on performance, cost and adaptability. Dedicated hardware wins where latency and resilience matter most. Software and cloud capacity earn their place with new services, regional productions and temporary peaks.

His rubric is useful because it works one workload at a time. A team that has treated the cloud as an all-or-nothing migration can run its own signal chain through it. Transcoding and graphics land on the cloud side quickly, and intercom and IFB land on the venue side. The argument then narrows to the few items that sit right on the line.

The link to the venue decides it

Riedel Communications sees the same split, and it puts the friction at the venue. The connection back to on-site talent, cameras, microphones, tally and IFB can still undermine the production or the business case. Every item on that list either involves a person who needs a cue right now or feeds the gear that sends it.

In Dillon's analysis, dividing workloads into elastic and sticky groups suggests hybrid production is probably where things settle for good, not a stop on the way to something else. If that reading holds, it changes procurement. On-site hardware for latency-bound work becomes a lasting part of the facility, and buying it doesn't mean betting against the cloud. In his view the industry largely knows which work belongs in the cloud and isn't confused about it. The hard part is connecting the elastic and sticky pieces so that teams don't end up running two separate operational worlds.

Integration moved into software

Interoperability used to mean getting one vendor's signal into another vendor's system without a conversion box. Much of that plumbing is now standard: SRT, NDI, JPEG XS, SMPTE ST 2110 and NMOS cover a lot of what used to need custom work.

Wagdin says the problems that remain sit at the boundaries between systems. Control models, protocols, timing implementations and media formats all meet there, and they still need to work together predictably. NMOS and ST 2110 help, but orchestration and end-to-end testing are still part of making a multi-vendor environment function as one system.

Ulrich Voigt, director of live production solutions at Riedel, takes it a level higher. For him the tougher problem is coordinating the lifecycle of the multi-vendor infrastructure itself: provisioning and deprovisioning services, assigning resources, applying configurations and tracking dependencies across vendors. Containerization helps, but Voigt notes that not every broadcast application containerizes efficiently yet.

Put together, the two accounts suggest that integration labor hasn't gone away. It has changed shape. A fault that once meant swapping a box in the rack now means a configuration someone has to apply, verify and later tear down across several vendors' software, often on both sides of the cloud-to-venue line. Dillon calls this a move from a hardware-integration problem the industry understood to a software-orchestration problem it is still learning to manage.

One event, many outputs

Each of those versions sits on the elastic side of the line: graphics, packaging, transcoding. That suggests the multi-output model fits the cloud well. It also points back to Voigt's problem. Every added destination is another set of services to provision, configure and monitor, so the orchestration load grows with the number of outputs even when compute is easy to get. For operators whose calendars swing between a few tentpole events and quieter weeks, those temporary peaks are exactly where Wagdin says cloud capacity earns its keep.

There's a fair objection to calling hybrid permanent. When Voigt says "yet" about containerization, he implies some of today's limits will ease. If audio mixing or intercom software comes to run efficiently in containers with acceptable latency, the line moves, and the hardware budget moves with it. Still, talent, cameras and microphones are at the venue, and better software doesn't change that. What Riedel describes is mostly physics and presence.

For operators planning hybrid productions, this suggests the seam between the cloud half and the on-site half could keep taking engineering time and should be part of budget planning.