Introduction
Latency, the delay between something happening in front of a camera and the moment it shows up on a screen, is a serious concern across video production: live broadcasts, IMAG for concerts, esports, virtual production, LED wall work, motion capture, live streaming, interactive productions, and camera operating all depend on it. Even a few extra milliseconds can break the illusion, the timing, or an operator’s ability to actually do their job. Latency isn’t caused by any single device, either; it accumulates across an entire pipeline. This article breaks down every source of delay, how they stack together, how different technologies compare, and how to actually design a low-latency workflow.
What Is Latency?
Latency is the time between input and output, measured in milliseconds or in frames of delay (one frame at 60fps works out to roughly 16.67 ms). It builds up across the whole chain: capture, encode, transport, process, and finally display.
Why Latency Matters
If a monitor lags, a camera operator can’t accurately pull focus, pan, or track movement. In IMAG setups, if the screens in a venue show performers out of sync with their real movements, the audience notices immediately. In virtual production, latency breaks the whole illusion: camera tracking desyncs, LED walls lag behind movement, and Unreal Engine can’t match perspective in real time. In live streaming, too much delay makes real interaction awkward or impossible. In esports, high latency makes viewing difficult and undermines competitive integrity outright.
Camera Sensor & Internal Processing
A camera adds latency before the signal even leaves the body, caused by rolling shutter readout, debayering, noise reduction, image scaling, Log/HDR processing, and internal frame synchronizers.
| Camera Type | Latency |
|---|---|
| Cinema cameras (ARRI, RED, Sony Cine) | 1–2 frames |
| Mirrorless cameras | 1–3 frames |
| Broadcast cameras | <1 frame |
| Smartphones | 3–10 frames |
Mirrorless cameras often run high on latency simply because they prioritize image processing quality over real-time response.
Transmission Medium: SDI vs HDMI vs IP Video
SDI latency is virtually zero, typically under a microsecond, since it’s an uncompressed electrical digital signal with no buffering. HDMI is also near-zero for raw transmission, though handshakes, color space conversions, and EDID negotiations can add slight delays on certain devices. IP video is where latency really starts to grow:
| Protocol | Latency |
|---|---|
| NDI Full Bandwidth | 1–2 frames |
| NDI HX3 | 20–50 ms |
| NDI HX | 50–200 ms |
| SRT | 120–2,000+ ms |
| RTMP | 2–5 seconds |
| WebRTC | 100–500 ms |
IP video is genuinely powerful, but not every format suits a low-latency environment.
Switchers, Scalers, and Converters
Nearly every device that touches a signal adds some frame sync delay.
| Device Type | Latency |
|---|---|
| Hardware switchers (ATEM, TriCaster) | 0–1 frame |
| Video scalers | 1–3 frames |
| Frame synchronizers | 1–2 frames |
| Cross-converters | 0–1 frame (SDI↔HDMI) |
| Wireless SDI transmitters | 1–4 frames |
| Wireless HDMI transmitters | 3–10 frames |
Wireless is consistently the biggest offender here, since it requires encoding, packetization, RF transmission, and decoding all in sequence, and that latency compounds fast.
Encoding & Streaming
Encoding compresses video for streaming, recording, network transport, or wireless systems, and heavier compression means more latency.
| Codec | Latency |
|---|---|
| ProRes / DNx | 1–2 frames |
| H.264 | 1–4 frames |
| H.265 | 2–8 frames |
| AV1 | 10–50+ frames (high computational cost) |
Hardware encoders consistently beat software encoders on latency by a wide margin.
Displays & Monitors
Monitors add their own delay through scaling, color processing, HDR tone-mapping, refresh synchronization, and overdrive or frame interpolation.
| Display Type | Latency |
|---|---|
| Professional SDI monitors (SmallHD, Flanders) | <1 frame |
| Broadcast studio monitors | 1–2 frames |
| Consumer TVs | 3–20 frames |
| Gaming monitors | 1–3 frames |
| LED walls | 2–6 frames |
LED walls specifically add delay because the processor has to handle scaling, color calibration, HDR mapping, refresh driving, and module timing all at once.
Cumulative Latency: How It Adds Up
Latency stacks linearly through the whole pipeline. Take a camera at 2 frames, a wireless transmitter at 3, a switcher at 1, an LED processor at 3, and an LED wall at 2: that totals 11 frames, roughly 183 ms at 60fps. That’s genuinely visible to the human eye and will break timing for both IMAG and virtual production.
Latency Thresholds: What Is “Acceptable”?
| Application | Acceptable Latency |
|---|---|
| Virtual production (LED volumes) | <5 ms |
| Camera operating | <1 frame |
| IMAG (concerts) | <2–3 frames |
| Live broadcast switching | <3 frames |
| Esports / gaming | <10 ms |
| Streaming to internet | <5 seconds |
| Remote contribution | <200 ms |
Virtual production carries the strictest requirement of all; even a single dropped frame can shatter the illusion.
How to Build a Low-Latency Pipeline
A few engineering rules hold up consistently. Use SDI whenever possible: no compression, no network buffering, no jitter, and the lowest latency transmission available, with better cable runs and locking connectors on top, which is why SDI wins out over HDMI in any serious production. Avoid wireless unless it’s genuinely necessary, and if it is, stick with professional zero-delay SDI systems like the Teradek Bolt series or Hollyland Mars 4K rather than consumer HDMI transmitters. Minimize frame synchronizers, since anything reconciling mismatched signals adds 1 to 2 frames, and matching camera frame rates up front avoids that entirely. Avoid unnecessary scaling too; converting between 1080p, 4K, and 720p is a hidden latency cost, so keeping resolutions consistent pays off. For IP video, use NDI Full Bandwidth rather than NDI HX when low latency actually matters, saving HX for situations where bandwidth is genuinely limited. Use proper monitors, since professional SDI monitors from Flanders, SmallHD, or Sony run very low latency while consumer TVs are consistently the worst offenders. And LED walls need special care: virtual production pipelines need genlock, low-latency processors, and frame-accurate sync, since most LED walls default to 2 to 6 frames of delay; processors from Brompton or Megapixel VR tend to give the best results here.
Genlock & Synchronization
Latency isn’t the only thing that matters; sync matters just as much. If camera, processor, LED wall, and display aren’t all perfectly in sync, the result is judder, rolling bars, mismatched motion, or Unreal Engine tracking errors. Genlock is what keeps every system running on the same timing clock.
Measuring Latency
Latency can be tested with a stopwatch app filmed by the camera itself, a timecode generator, the LED flash method, software latency meters, tools like the Sync-One2 analyzer, or the internal latency measurements built into cameras like Sony Venice or RED. Most of these methods just capture a timestamp on both ends of the pipeline and compare.
Latency Optimization Checklist
Use SDI as the top priority whenever possible. Avoid unnecessary conversions, no HDMI-to-SDI-to-HDMI-to-SDI loops. Avoid wireless when it can be avoided, since even the best systems add 1 to 4 frames. Use professional equipment: real LED processors, real SDI monitors, hardware encoders. Keep formats consistent: same frame rate, same resolution, same time base throughout. Use genlock, especially for virtual production and LED wall work. And upgrade the switchgear, high-bandwidth SDI routers for SDI workflows, multicast-enabled, low-buffer network switches for IP.
Latency by Workflow Type
Cinema production tolerates under 2 frames, typically SDI into hardware wireless into SDI monitors. IMAG and live events tolerate 3 to 5 frames, running SDI through a switcher into an LED processor. Esports needs under 10 ms, relying on gaming monitors with direct SDI or HDMI feeds. Live broadcast tolerates under 3 frames through an SDI router, switcher, and broadcast monitors. Streaming tolerates 1 to 5 seconds using SRT, RTMP, or a hardware encoder. Virtual production with an LED volume needs under 3 to 5 ms end-to-end, running SDI through genlocked cameras into Brompton or Megapixel processors.
Conclusion
Latency is one of the most critical technical considerations in video production, and also one of the least understood. Every component in a pipeline adds its share of delay, but with the right design choices it’s entirely possible to build an ultra-low-latency workflow suited to anything from basic streaming to high-end virtual production. SDI offers the lowest latency, HDMI sits at a moderate level, NDI Full Bandwidth stays low while NDI HX and HX3 run medium, LED walls run high unless carefully optimized, wireless links always add delay, scaling and sync issues are hidden latency killers, and genlock is essentially mandatory for virtual production and LED work. Understanding exactly where latency comes from is what lets cinematographers, engineers, and streaming professionals build systems that stay fast, responsive, and true to the timing and realism the work actually demands.