Read in
17
min
REMI Production: What Teams Get Wrong About Connectivity
Short answer: REMI (remote integration model) production sends camera, audio and data feeds from the venue over IP to a central control room, so every feed rides the venue uplink. MR·NET, a managed bonded cellular service from MOTIONRAY Inc., bonds Verizon, AT&T and T-Mobile into one connection for contribution feeds, return video and intercom.
What is REMI production?
REMI stands for remote integration model. The cameras, microphones and a small crew stay at the venue. The director, switcher, graphics, replay and audio mix sit in a production center somewhere else. Every camera feed, every mic and every intercom channel travels between the two over IP.
That’s the whole appeal: fewer people on the road, one control room serving many events, and the same senior crew covering more shows in a week. It’s also the whole risk. In a traditional production, a bad internet connection ruins the stream to viewers. In REMI, a bad connection ruins the production itself, because the director can’t see the cameras.
How is REMI different from a traditional outside broadcast?
Traditional outside broadcast | REMI production | |
|---|---|---|
Where the control room is | In a truck or flypack at the venue | At a central production center |
Crew on site | Full crew: director, switcher, audio, graphics, replay, camera ops | Camera ops and a technical lead |
What leaves the venue | One finished program feed | Every camera, audio and data feed |
What the connection carries | One stream out | Many streams out, return video and intercom both ways |
What a dropped connection breaks | Viewers lose the program | The director loses the cameras |
How is REMI different from cloud production?
In REMI, the feeds go to a physical control room the team owns or rents. In cloud production, the feeds go to switching, graphics and encoding software running on cloud servers, and the crew operates it from wherever they are. The connectivity problem at the venue is the same in both: every feed still has to leave the building over the uplink.
Why is REMI a systems integration problem?
Most REMI planning starts with the video chain: which cameras, which encoders, which transport protocol, which switcher at home. Those are solved problems with good vendors behind them.
REMI usually fails one layer down, at connectivity. The encoder can be flawless and the protocol can recover lost packets, but neither can create upload that isn’t there. A REMI workflow is only as good as the network under it, and the network at a venue is the part the production team controls least.
The gaps show up between systems. Video contribution usually gets engineered carefully: encoders specified, codecs tuned, SRT settings configured. Intercom, tally and camera control get treated as an afterthought, even though in many productions they are more sensitive to delay than the video itself. A camera operator who hears the director late can’t take direction cleanly. If the video path had a spec and the intercom ran over whatever the venue Wi-Fi offered, nobody owned the part that failed.
Treat connectivity as part of the system you integrate, with the same rigor as the video chain: specified, tested, redundant and monitored, for every signal path, not just the video feed.
Why does best-effort internet break REMI?
Most venue internet is best-effort. It promises to try, not to deliver a number. Venue Wi-Fi is shared with attendees, staff, vendors and exhibitors. A shared wired drop can be shared too, behind the same venue firewall.
Best-effort is fine for email. It isn’t fine for six camera feeds that the director needs in real time. The trouble is timing: best-effort networks look excellent during load-in, when the building is empty, and get worse as the doors open and the crowd arrives, right when the show starts.
If the venue offers a dedicated line with a guaranteed speed, take it seriously. Ask for the upload figure in writing, ask whether it is dedicated or shared, and still plan a second path.
Then audit what you’re actually getting. A spec sheet number is not a measurement. Ask for a packet loss test, a round-trip time measurement and a jitter sample, ideally taken during a real event, not on an empty Tuesday.
The 5G reality check: is 5G better than 4G for REMI?
Not necessarily. Much of what’s sold as 5G today is non-standalone (NSA) 5G: the radio is 5G, but the core network behind it is still 4G LTE. The headline speed gains are mostly a download story, and REMI is almost all upload. Higher-frequency 5G bands also get into buildings less reliably than the bands LTE uses.
In MR·NET’s own deployments, LTE has given steadier upload in crowded indoor venues than 5G NSA. And whatever the generation, a crowded venue puts thousands of phones on the same nearby towers, and any single connection is still a single point of failure.
What matters more than the generation of one modem:
Carrier diversity. Different carriers run different towers and different backhaul. When one gets busy, the others may not.
Plan tier. Enterprise-tier SIMs are a separate lever that helps curb congestion in a crowded venue.
Bonding. All carriers working at once, so one slow link doesn’t stall the feeds.
For clarity: MR·NET’s gateway uses three LTE modems, one each on Verizon, AT&T and T-Mobile, on high-priority enterprise-tier SIMs. MR·NET’s case rests on carrier diversity, plan tier and bonding, not on a modem generation.
Why does single-path connectivity fail?
A single path is one venue line, one cellular modem, or one carrier. It fails in ways you can’t control from the production center: a cut cable, a misconfigured venue firewall, a congested tower, a carrier outage in that area.
Two connections from the same source can fail together. Two modems on the same carrier share the same towers. Two venue drops share the same venue uplink. Real redundancy means paths that fail for different reasons.
For REMI, failover isn’t enough either. Failover switches to a backup after the primary dies, and the switch itself can break live feeds. Bonding keeps every path working at once, so a failing link doesn’t take the feeds down with it. Read how many carriers to bond at a crowded venue.
The broadcast industry wrote this lesson into a standard long ago: SMPTE ST 2022-7 covers hitless failover between redundant IP paths. Smaller productions still skip the second path because it costs money and “the line has never failed before.” It fails eventually. The only question is whether it happens in rehearsal or on air.
How much upload does a REMI production need?
Add up every feed that leaves the venue at once, then multiply by 1.5 for headroom. Return video and program feeds coming back to the venue use download, so size them separately.
Step 1: list what goes up.
Feed leaving the venue | Your bitrate |
|---|---|
Each camera feed | Set by your encoder settings |
Embedded or separate audio | From your audio system |
Intercom and talkback | From your intercom system’s documentation |
Data (graphics data, scoring, telemetry) | Usually small; check your systems |
Camera control or remote operation | Check your systems |
Step 2: multiply the total by 1.5.
Two worked examples (bitrates are illustrative; use your own encoder settings):
Production | Feeds up | Total sent at once | Upload to plan for (x1.5) | Data per hour | MR·NET data cost per hour at $8/GB |
|---|---|---|---|---|---|
Small REMI: 3 cameras at 6 Mbps | 3 | 18 Mbps | About 27 Mbps | About 9.7 GB | About $78 |
Larger REMI: 4 cameras at 14 Mbps | 4 | 56 Mbps | About 84 Mbps | About 30 GB | About $242 |
Data uses the rule of about 0.54 GB per hour per 1 Mbps, including about 20% overhead. Add intercom and data feeds to both rows.
The MR·NET MR 006 Plus supports up to 250 Mbps maximum bandwidth, but real throughput at a venue depends on carrier signal at your position and how many carriers are active. Test at the production position, at show time, before you commit to a camera count. For single-stream targets, see how much upload speed you need for live streaming.
SIM banks vs managed bonded cellular: what’s the difference?
Some teams build their own cellular setup: a router with a bank of SIM cards, bought on whatever plans they can get, and someone on the crew who knows how to configure it.
It can work. SIM banks are inexpensive and useful for low-stakes jobs where the main need is more than one cellular path. For REMI they leave several jobs with the production team:
Switching vs bonding. Many multi-SIM routers switch between carriers rather than bonding them. One carrier at a time is failover, not bonding.
Plan tier. Carrier count and plan tier are separate levers. Enterprise-tier SIMs help curb congestion in a crowded venue; a bank of consumer SIMs gives you carrier count without that second lever.
The server side. True bonding needs a bonding server to rejoin traffic. Someone has to run it.
Monitoring and support. When a link degrades mid-show, someone has to see it and act.
A managed bonded cellular service puts those jobs in one plan. MR·NET ships a preconfigured gateway with high-priority enterprise-tier SIMs on Verizon, AT&T and T-Mobile, runs the cloud bonding server, and the connection is cloud managed and monitored by MR·NET’s system and support team. MR·NET’s US-based support answers by chat or phone, after hours too; MR·NET says its average response time is under a minute. Compare the options in managed bonded cellular vs buying a router vs renting a kit.
The choice comes down to in-house network skills and volume. For a production company running 30 or 40 events a year, the DIY route piles up work: carrier plans, hardware configuration, and field troubleshooting three hours before show time. DIY works with the right skills at modest scale. At volume, managed usually wins once you count that overhead honestly.
What does a proper REMI connectivity architecture look like?
Design the network the way you design the video chain: layers, each able to carry the show if another fails.
Layer 1: the bonded network at the venue
One bonding gateway serves the whole production. Encoders, intercom, laptops and camera control plug into its Ethernet ports or join its Wi-Fi. The MR·NET MR 006 Plus has four Gigabit Ethernet ports and Wi-Fi 6, and bonds three carriers. Where the venue offers a wired line, add it to the bond as another source instead of running it separately: the MR 006 Plus supports up to six sources, including wired WAN, local Wi-Fi and satellite.
Inside concrete, steel or multi-level buildings, plan external antennas. The antenna moves the reception point outside the structure while the gateway stays at the production position. MR·NET’s custom plan lists advanced antennas as an option. Where you can, put video contribution, intercom, tally and return feeds on separate network segments, so a congested contribution path doesn’t take down director-to-camera comms.
Layer 2: bonded encoders on top, where they fit
A bonded encoder is built to move one video feed over several paths. A bonded encoder with Ethernet or Wi-Fi can run over MR·NET as one of its paths, so a camera feed gets the encoder’s own resilience plus the bonded network underneath it. Read bonding router vs bonded encoder.
Layer 3: a return path you’ve sized
Return video, program feeds for monitors and the director’s talkback come back down to the venue. Size download for them, and test them along with the upload.
Layer 4: monitoring that someone actually watches
Assign one person at the production center to watch connection health, not just the pictures. MR·NET’s management portal shows connection status and data use per gateway.
Layer 5: a tested fallback plan
Decide in advance what you drop first if capacity falls: a lower bitrate on secondary cameras, then fewer iso feeds, then a single program feed. Rehearse it.
Three tests make the plan real:
A carrier-by-carrier survey. Big-market venues often have strong indoor coverage on all three carriers. Smaller-city arenas and unusual buildings need a signal check per carrier at the actual production position, not just the venue address.
A pull test in rehearsal. Disconnect one path with sessions running and confirm the rest carry the show.
A load test under show conditions. Run every contribution path at once before the doors open. A quiet venue doesn’t predict a full one.
What does the venue need, and what does the production center need?
At the venue | At the production center | |
|---|---|---|
Video | Cameras, encoders (or bonded encoders), embedders | Decoders or receivers, switcher, replay, graphics |
Audio | Mics, audio embedding or a separate audio path | Audio mixing |
Communications | Intercom endpoints for camera ops and the on-site lead | Intercom matrix, director and producer positions |
Network | Uplink with enough upload for every feed x1.5, ideally bonded across carriers, plus any venue line in the same bond | A stable, wired connection with enough download for every incoming feed and enough upload for returns |
Power | Power for every device, including the gateway (the MR 006 Plus runs on 12 to 54 V DC at about 24 W) | Standard facility power with backup |
People | Camera operators and one technical lead who owns the on-site network | Director, operators, and one person watching network health |
What infrastructure is required at the venue and production center for a REMI setup?
At the venue: cameras and encoders, audio capture, intercom endpoints, power, and an uplink sized at about 1.5 times every feed you send at once, ideally bonded across several carriers. At the production center: receivers or decoders, switching, audio, graphics, an intercom matrix, and a stable wired connection with enough download for every incoming feed.
How do I spec the contribution chain for a multi-site REMI workflow?
Spec each venue as its own chain, then standardize. For every site: list each feed and its bitrate, multiply by 1.5 for the upload target, choose the transport protocol and its latency setting, define the primary path (bonded cellular, with any venue line in the bond), and name who monitors it. Use identical kits and settings at every site, so the production center sees the same thing from each.
What latency and reliability should REMI contribution meet?
There is no single industry number. The right target depends on what the production does with the feeds:
Camera feeds the director cuts live need low, steady latency across all cameras. Consistency matters as much as the number: feeds that arrive at different delays are hard to cut.
Intercom needs the lowest delay you can get, because people talk over each other when it lags.
Program contribution to a platform can tolerate more delay in exchange for resilience.
Transport protocols such as SRT handle packet loss by holding a short buffer and resending what goes missing. A bigger buffer recovers more loss but adds delay; a smaller one does the opposite. Set it from measured conditions on the actual connection, at show time.
Haivision gives a starting formula in its guide to configuring SRT settings (checked October 2026): SRT latency equals an RTT multiplier times the round-trip time (RTT). The multiplier ranges from 3 to 20. Below 3 is too small for SRT to be effective, and Haivision says anything above 20 implies a network with 100% packet loss. If the measured RTT is 20 ms or less, Haivision uses 20 ms as the RTT value. Measure RTT on the bonded connection at the production position, then pick the multiplier from the loss you see there.
On reliability, set the target in operational terms: no lost camera feed during the program, and a rehearsed fallback if capacity drops. Bonding three carriers is how MR·NET approaches that target: traffic can move to the remaining carriers if one slows or drops, though available speed can dip while fewer are active.
Which connection should carry REMI contribution?
Option | Paths | Carries | Plan tier | When one path fails | The catch |
|---|---|---|---|---|---|
Venue hardline | One | Everything plugged into it | Venue’s network | Everything stops | Shared or best-effort unless dedicated; price and deadlines vary by venue |
Single-carrier cellular | One carrier | Everything on that modem | Often consumer | Everything stops | Shares towers with the crowd |
Bonded encoder | Several, for one feed | One video feed | Varies by unit | That feed can continue on other paths | Covers one feed, not intercom, laptops or other cameras |
MR·NET bonded network | Verizon, AT&T and T-Mobile at once, plus optional wired, Wi-Fi or satellite sources | Every device: encoders, intercom, laptops, over 4 GbE ports and Wi-Fi 6 | High-priority enterprise-tier SIMs | Traffic can move to the remaining carriers | Needs carrier signal; speed can dip when fewer carriers are active |
The strongest setups combine the last two: bonded encoders for critical camera feeds, running over the MR·NET bonded network that also carries everything else. See bonded cellular for broadcast and REMI.
Which vendors do REMI teams work with?
On the video chain, makers such as LiveU, Haivision, Matrox and Ross Video publish REMI workflows and guides on their own sites (checked October 2026). Most of that material focuses on the video chain: encoders, transport, switching.
MR·NET works one layer down, on the network at the venue. A bonded encoder with Ethernet or Wi-Fi can run over MR·NET as one of its paths, so teams don’t have to choose between the two.
What’s the pattern behind REMI failures?
Look at REMI shows that went wrong and the pattern repeats: the video chain was specified in detail, and the connection was assumed. Someone tested the venue internet during load-in, it looked fine, and it fell over when the crowd arrived.
The teams that run REMI week after week flip that. They treat the uplink as part of the production, size it at 1.5 times every feed, bond carriers that fail for different reasons, put a person on monitoring, and rehearse the fallback. Repeatable connectivity is what lets a production scale. North Texas streamer tr3dio has about 2,200 past live streams on its YouTube channel (as of October 28, 2026), and MR·NET says about 95% of the channel’s streams run through MR·NET. Read how tr3dio streams 2,000+ North Texas games with MR·NET.
The technology keeps improving. The connectivity mistakes stay the same. More on that in network reliability in modern video productions.
Sources: Haivision, “How to configure SRT settings on your video encoder for optimal performance” (checked October 2026); SMPTE ST 2022-7; MR·NET MR 006 Plus specifications and pricing (mrnet.us, checked October 2026); tr3dio YouTube channel, observed October 28, 2026.
Frequently Asked Questions
What is REMI production?
REMI, the remote integration model, keeps cameras and a small crew at the venue and moves the control room to a central production center. Camera, audio and data feeds travel from the venue over IP, and return video and intercom travel back. That puts the whole production on the venue uplink.
What are REMI production’s internet requirements?
Enough upload for every feed you send at once, multiplied by 1.5 for headroom, on paths that fail for different reasons. The total depends on stream count and codec, so add up your own encoder settings. MR·NET bonds Verizon, AT&T and T-Mobile into one connection on high-priority enterprise-tier SIMs, and a venue line can be added to the same bond as another source.
How do you build reliable connectivity for remote production?
Bond several carriers at once rather than switching between SIMs, on enterprise-tier plans, with someone watching link health during the show. Engineer intercom, tally and camera control paths with the same care as video, since those are the paths that usually get treated as an afterthought. Then test it: a pull test in rehearsal and a load test before doors open.
Is cellular bonding good enough for REMI broadcast?
Often, where carriers have good signal at the production position, and if it is true multi-carrier bonding rather than a basic SIM bank. Traffic can move to the remaining carriers if one slows or drops, though speed can dip while fewer are active. Test at the venue at show time and size your camera count to what the connection holds.
How much upload does a REMI production need?
Add every feed leaving the venue, then multiply by 1.5. Three camera feeds at 6 Mbps total 18 Mbps, so plan about 27 Mbps. Four feeds at 14 Mbps total 56 Mbps, so plan about 84 Mbps. Add intercom and data feeds on top.
Is 5G better than 4G for live production contribution?
Not necessarily. Much of today’s 5G is non-standalone: a 5G radio on a 4G LTE core, and its speed gains are mostly on download, while contribution is upload. In MR·NET’s own deployments, LTE has given steadier upload in crowded indoor venues. Carrier diversity, plan tier and bonding matter more than the generation of one modem.
What’s the difference between a SIM bank and managed bonded cellular?
A SIM bank holds several SIM cards and spreads traffic between them, often switching carriers rather than bonding them, and you manage the plans and fix it yourself. Managed bonded cellular uses all carriers at once through a bonding server, keeps sessions alive when a link drops, and comes with SIMs, monitoring and support in one plan. MR·NET uses enterprise-tier Verizon, AT&T and T-Mobile SIMs.
Why do REMI productions fail even with good equipment?
Most REMI failures are systems integration failures, not equipment failures. Teams engineer video contribution carefully but run intercom, tally and camera control on an untested path, test the venue connection only while the building is empty, or rely on a single contribution path with no tested failover. SMPTE ST 2022-7 exists for hitless failover between redundant IP paths.
Why not just use a bonded encoder for REMI?
A bonded encoder protects one video feed. REMI also needs intercom, return video, laptops and other cameras online. A bonded encoder with Ethernet or Wi-Fi can run over MR·NET as one of its paths, so the critical feed gets both layers of protection.
What is the difference between REMI and cloud production?
In REMI, venue feeds go to a physical control room. In cloud production, they go to switching and graphics software running on cloud servers that the crew operates remotely. Both put every feed on the venue uplink, so the connectivity plan is the same.
What equipment is needed at the venue for REMI?
Cameras, encoders or bonded encoders, audio capture, intercom endpoints, power, and a network uplink sized at about 1.5 times every feed you send at once. A bonding gateway such as the MR·NET MR 006 Plus connects them over four Gigabit Ethernet ports and Wi-Fi 6.
