Quick Answer
If the offset is fixed, you can usually correct it with a measured audio or video delay. If the offset changes over time, a static delay will not solve it. In that case, you need to fix the underlying timing, buffering, or network problem first.
What Is AV Sync Error?
Audio/video sync error — commonly called lip sync error — occurs when the audio and corresponding picture arrive at different times. Perception depends on programme content, viewing conditions, and whether sound leads or follows the picture, so a single “visible at X ms” number is not a universal diagnostic threshold.
The paths can diverge for many reasons: frame synchronization, scaling, colour conversion, encoding, decoding, audio processing, network buffering, clock drift, or monitoring. Video is often the slower path, but that is a measurement result to establish, not a rule to assume.
EBU lip-sync guidance:
EBU R37 recommends keeping each production stage within 5 ms sound early to 15 ms sound late. For an output intended for emission, its overall limits are 40 ms sound early to 60 ms sound late. These are limits, not a reason to stop improving a measurable offset.
Why It Happens
AV sync problems usually come from one of two patterns: a fixed offset or a variable offset. Fixed offsets are caused by predictable latency in one path. Variable offsets come from jitter, unstable timing, dynamic buffering, or load-dependent processing.
The distinction matters. A fixed 40 ms error can be corrected. A drifting 40 ms error needs root-cause work before any compensation is trustworthy.
Step 1: Measure Your Actual Offset
Before applying any compensation, measure the actual offset in your system. Guessing and adjusting by ear is unreliable. Use one of these methods:
Method A: Camera + Clap or Flash Reference
Use a clapper, hand clap, slate, or a short flash-and-beep reference that you can send through the real chain you want to measure. The useful number is the offset you can see or hear in the actual output path, not a browser-only estimate. Record the millisecond reading you observe in the captured output.
Method B: Clapperboard Reference
Film a traditional clapperboard on the camera and feed it through your production chain. In editing software, scrub frame-by-frame to find the frame where the clapper hits, then align it with the audio transient. The frame difference × (1000 ÷ frame rate in fps) = offset in ms.
Method C: Purpose-built measurement
For an electrical or automated test, use a documented flash-and-blip source and a measurement system that captures both output components on the same timebase. State the measurement point and resolution; do not claim sub-millisecond accuracy from a frame-rate-limited recording.
Step 2: Identify the Source of Delay
Once you know the magnitude and direction of the offset, identify where the paths diverge. “Audio early” describes the measured relationship; it does not by itself prove which component caused the delay. Common contributors include:
- Capture card processing: Capture devices can add meaningful latency, but the range varies widely by model, buffering mode, resolution, and driver path.
- Video software processing: vMix, OBS Studio, and similar systems add latency through their render and compositing pipelines. The real value depends on load, buffering, and output path.
- NDI High Bandwidth: Usually low-latency, but still non-zero and implementation-dependent.
- NDI HX3 / HX2: Usually adds more delay than NDI High Bandwidth, with the exact amount depending on device implementation and compression strategy.
- Dante / AES67 audio network: Compensation depends on the devices involved, the path design, and where the audio is finally received or mixed.
- Audio interface buffer: Buffer size still matters, and multiple processing stages can stack into a meaningful offset.
Step 3: Apply Compensation
There are two compensation strategies: delay the faster signal to match the slower one, or reduce the latency of the slower path. In live production, reducing latency is often preferable since adding delay worsens the overall end-to-end delay for the operator. However, for broadcast output the critical requirement is sync, not minimum latency.
Fixing Sync in OBS Studio
- Open the Audio Mixer panel.
- Click the gear icon (⚙) next to the audio source that is early.
- Select Advanced Audio Properties.
- In the Sync Offset column, enter a positive value in milliseconds to delay the audio.
- If audio is late (video is early), add OBS's Render Delay filter to the affected video source where appropriate. It delays that source's rendering, not every output globally, and does not correct drift. Measure the actual recorded or returned output after adjustment; if the source filter cannot cover the required path or delay, review audio processing latency or use an appropriate external video-delay device.
Typical range: common offset values vary a lot by capture chain, GPU load, transport method, and monitoring path. Use measured offsets as the source of truth instead of copying a published example.
Fixing Sync in vMix
- Open the Audio Mixer and select the settings button for the affected audio input.
- In the input's audio settings, use Delay to add the measured value in milliseconds.
- Record or monitor the real output and verify the result.
- Use default or output delay settings only when the whole path has been measured and needs the same treatment.
Fixing Sync in Other Production Systems
On a switcher, replay system, or production appliance, find the documented input or output delay control for the exact software and hardware version. Some controls work in milliseconds and others in frames. Confirm whether the setting delays audio, video, or the complete output before applying it.
Fixing Sync in Dante Networks
Dante latency settings and compensation behavior depend on the devices involved and where the audio is received or mixed. When audio travels over Dante while video takes a different, longer path:
- Verify the relevant device latency settings and the actual audio path first.
- Decide whether the needed compensation belongs in the networked audio device, the destination device, or the downstream mixer/application.
- Use latency settings that match the stability requirement of the network path rather than treating one published value as universally correct.
In real systems, different receivers may need different treatment. Do not assume that one Dante setting solves every destination in the chain.
Step 4: Verify in the Real Output Path
After applying compensation, verify the result with a real recording or programme-output capture. Aim to minimize the measured difference and use the correct stage or end-to-end R37 range for the point being assessed. Re-check after any change to the signal chain.
Persistent or Variable Sync Drift
If the offset changes over time, static compensation is not enough. The cause may be clock-rate drift, changing buffers, variable processing time, packet-delay variation, or a combination of them; “jitter” is not a complete diagnosis. Common causes include:
- Network congestion causing variable NDI or Dante packet delivery timing
- Clock or sample-rate mismatch, including an incorrect or unstable network-clock design where the devices actually depend on one
- GPU render time variation causing frame latency fluctuation in software mixers
- Temperature-driven clock drift in non-genlocked systems over multi-hour productions
For a changing offset, address the measured cause — clocking, network behavior, processing load, or buffering — rather than applying one static correction.
Common Mistakes
- Applying delay before measuring the real offset.
- Treating variable drift like a fixed offset problem.
- Assuming a named product always adds the same number of milliseconds.
- Fixing one destination while ignoring the rest of the production chain.
Related Tools & Next Step
References