Render Speed Check-In

[Dave Labrecque]

Today I did an hour-long screen capture of audio and video of a thing for my wife (using open-source OBS Studio--pretty sweet!). She also wanted the audio as a separate file, so I spun off the audio into a WAV file using VLC Media Player. That extraction took a few seconds.

Then I placed that audio on a track in SAWStudio64 V2, did maybe four edits, taking out maybe a second or two, total. Then I did a BuildMix to a new WAV file. That took, like, three minutes to render. Is this indicative of something on my system that needs tweaking? Thoughts?

The original audio was AAC inside an MP4 of some kind. 48KHz was the original audio sample rate, which I maintained through the entire process.

Is the explanation as simple as: the AAC audio, being highly compressed, can be read faster than WAV audio can be? Not sure if that makes sense, because both the extraction and the BuildMix also wrote WAV audio, which is slower than reading. Right? :confused:

[cgrafx]

Today I did an hour-long screen capture of audio and video of a thing for my wife (using open-source OBS Studio--pretty sweet!). She also wanted the audio as a separate file, so I spun off the audio into a WAV file using VLC Media Player. That extraction took a few seconds.

Then I placed that audio on a track in SAWStudio64 V2, did maybe four edits, taking out maybe a second or two, total. Then I did a BuildMix to a new WAV file. That took, like, three minutes to render. Is this indicative of something on my system that needs tweaking? Thoughts?

The original audio was AAC inside an MP4 of some kind. 48KHz was the original audio sample rate, which I maintained through the entire process.

Is the explanation as simple as: the AAC audio, being highly compressed, can be read faster than WAV audio can be? Not sure if that makes sense, because both the extraction and the BuildMix also wrote WAV audio, which is slower than reading. Right? :confused:

The audio extraction in VLC is fast because its not processing anything, simply writing the audio portion of the data to a separate new file

The BuildMix process in SAW is going to take longer because it actually has to process the audio, which it did as fast as your computer and the audio engine could process the data. So it handled processing an hour of audio in a 3 minutes.

Processing looks like this:
- Read data from the audio file, processing each data sample input to a channel in the mixer
- mix the audio (even though its a single channel its still a mix bus)
- process every audio sample out of the mix bus and write that data to a new file.

[Dave Labrecque]

The audio extraction in VLC is fast because its not processing anything, simply writing the audio portion of the data to a separate new file

The BuildMix process in SAW is going to take longer because it actually has to process the audio, which it did as fast as your computer and the audio engine could process the data. So it handled processing an hour of audio in a 3 minutes.

Processing looks like this:
- Read data from the audio file, processing each data sample input to a channel in the mixer
- mix the audio (even though its a single channel its still a mix bus)
- process every audio sample out of the mix bus and write that data to a new file.

So even though it's pretty much just shuffling the same bits from here to there (no other DSP of any kind) it takes that much longer? Because of a few more "stops" along the way? Each of which needs buffer cycles? Does it essentially boil down to something like the MP4-to-WAV extraction time multiplied by the number of "stops" (channel, mix bus, output file)? Or is it more complicated than that? I suppose there was no buffering for the extraction, for example.

[cgrafx]

So even though it's pretty much just shuffling the same bits from here to there (no other DSP of any kind) it takes that much longer? Because of a few more "stops" along the way? Each of which needs buffer cycles? Does it essentially boil down to something like the MP4-to-WAV extraction time multiplied by the number of "stops" (channel, mix bus, output file)? Or is it more complicated than that? I suppose there was no buffering for the extraction, for example.

SAW is not shuffling bits from here to there, its actually processing the audio, every sample in to every sample out. (this is not a file copy operation)

Editing audio in SAW is very different than just moving data unlike when VLC splits the audio from the video file. VLC doesn't have to do any audio processing, it literally is just moving data from one file to another, so for all practical purposes however long it takes to do a file copy.

[Dave Labrecque]

SAW is not shuffling bits from here to there, its actually processing the audio, every sample in to every sample out. (this is not a file copy operation)

Editing audio in SAW is very different than just moving data unlike when VLC splits the audio from the video file. VLC doesn't have to do any audio processing, it literally is just moving data from one file to another, so for all practical purposes however long it takes to do a file copy.

I guess I don't know what processing means. I figured if there are no changes needed, it would just push the data along. That's my level of expertise with this stuff.

[cgrafx]

I guess I don't know what processing means. I figured if there are no changes needed, it would just push the data along. That's my level of expertise with this stuff.

When you place an audio file onto the timeline in SAW Studio, you are asking the audio engine to process that audio. Doesn't matter if you haven't made any actual edits (changed the EQ, added FX, chopped up regions, etc). That data still has to be read from the disk, packetized into samples the audio engine can work on, then the audio engine sums every channel through the mixers master output bus(es) and then dumps that data from audio engine output buffer and gets written back to disk.

SAW does this very efficiently at faster than real-time (meaning if you drop a 1hr audio file into saw it doesn't take an hour to process the audio), but it still is limited by sample rate and buffer size constraints, because the data stream is quantized into packets (each of which has to be processed - operated on by the audio engine) and reassembled back into a data stream on the output.