When the voice of an input comes before or after the lips, you set this input's picture-sound offset (“Bild-Ton-Versatz”) on the desk. A positive value delays the sound, a negative value delays the picture. The range is −500 to +1000 ms.
The offset exists for every input separately. So you correct only the source where it is wrong.
Which direction you need
What comes too early is what gets delayed.
| What you observe | What to do | Value |
|---|---|---|
| The sound comes before the picture: you hear the clap before the hands meet. | Delay the sound | positive, for example +100 |
| The sound comes after the picture: the hands meet, the clap comes later. | Delay the picture | negative, for example −100 |
The more common case is sound that comes too early. Cameras and video processors often need longer for the picture than for the sound, above all when the microphone is connected separately from the camera.
Step by step
- Find the input. Cut to your scenes one after the other and note on which one the offset appears. If it appears on all of them alike, read the section on checking further down.
- Put the input on preview. Every change of the offset causes a short jump on this input. On preview nobody notices.
- Create a clear event. Clap your hands in front of the camera or knock two objects together. A hard, short sound with a visible movement is much easier to judge than speech.
- Set a first value. Begin with a coarse step, about 100 ms in the right direction.
- Check and refine. Repeat the clap. If it is better but not yet good, go further in the same direction. If it is now wrong the other way round, take half of it back. That gets you there in a few steps.
- Cut to the scene and check in the stream. Only there do you see what your viewers see.
What to judge the offset by
The picture on the desk is a control picture with few pictures per second. For the question whether a camera delivers a picture, that is enough. For an offset of a hundred milliseconds it is too coarse, also with “Flüssige Vorschau” (smooth preview).
The finished stream is what you can rely on. The easiest way is to send to a channel you can watch yourself as a trial, clap a few times and look at the result there. The stream reaches you with a few seconds of delay. That does not disturb the judging, because only the distance between picture and sound in the result matters.
When the value is not enough, or wanders
- The range is not enough. If you need more than 500 ms of picture delay or more than 1000 ms of sound delay, something fundamental is wrong at the source. Look there: many apps and encoders have a setting of their own for the sound delay.
- The offset changes in the course of the stream. A fixed value can only make up for a fixed offset. If the distance grows over time, the cause is at the source. Reconnect it and check its settings.
- The offset is different after restarting the app. Some sources do not always start picture and sound the same way. After every reconnect, check the offset briefly with a clap.
- All inputs seem offset. Make sure that you are really judging the stream and not the control picture in the browser together with “Mithören” (listen in).
What the offset does not change
The offset moves an input's picture and sound against each other. It does not make up for two cameras arriving in the studio at different times, for instance because they send with different SRT latency. For that you give the cameras the same latency. See Setting the SRT latency.