Stereotool shared library support!

Over the years, we’ve had a lot of requests to help integrate stereotool audio processing into liquidsoap scripts.

While we acknowledge that stereotool is a proprietary sotfware, we aim at being pragmatic and supporting our users so we are happy to announce that we now have full, integrated support for the shared library provided by Thimeo Audio Technologies. This should bring a much better experience when integrating tool to your script.

The following is an extract from the documentation. Please refer to the latest published one for an up-to date version!

Starting with version 2.2.0, liquidsoap supports the shared library distributed by Thimeo Audio Technology using the stereotool operator (and track.audio.stereotool for the low-level, track-specific equivalent).

This feature is enabled in all release builds of liquidsoap starting with rolling-release-v2.2.x and should be enabled if you compile liquidsoap with the optional ctypes-foreign opam module installed.

The operator can replace the use of the stereotool binary in your script and offers multiple benefits. In particular, it has a very low latency compared to using the binary and also operates synchronously.

The operator should be quite easy to use. Here’s an example:

# Define a source
s = ...

# Apply stereotool to it:
s = stereotool(
  library_file="/path/to/stereotool/shared/lib",
  license_key="my_license_key",
  preset="/path/to/preset/file",
  s
)

That’s it! You can apply as many stereotool operators as you wish and at any stage in the script, thanks to its synchronous nature. However, a current limitation is that the processed audio signal is slightly delayed.

This is because the operator has an internal processing buffer. We do plan on delaying metadata and track marks to match this latency but this has not yet been implemented and will probably have to wait for the 2.3.x release cycle.

This means that, until then, track switches and metadata updates might happen slightly earlier than the corresponding signal. We’re talking about 50ms to 100ms earlier, though, so that might not be a super big deal.

For the same reason, the source returned by stereotool is an audio-only source. Otherwise, other concurrent tracks such as video and etc would be slightly out of sync. If you need to use the operator in this kind of situation, you might want to use a ffmpeg filter to e.g. adjust the video’s PTS to match the audio delay.

In such case, you can refer to the latency method that is available on the source returned by the operator which should indicate the delay to compensate from the processed audio signal.

The operator’s preset parameter has a companion load_type parameter that can optionally be used to only load a subset of the preset. You might refer to the upstream documentation if you need to use it.

Lastly, stereotool is a proprietary software. While we actively promote open source, we also want to meet our users where they are and, for a lot of them, this means supporting the sound processing provided by the tool.

However, to use it, you will need a license. Using the operator without the proper license will not result in an error in your script but the audio signal might have spoken text and/or beeps added to it.

Using the operator with an invalid license will be reported in the logs. You might also use the valid_license method available on the source returned by the operator, which returns false if the license is invalid. In this case, the unlincensed_used_features method returns a string indicating which unlicensed features are being used.