Prevent Audio File Conversion During Soundbank Generation

Question

Is there a way to prevent Wwise from processing the input audio files through the conversion settings during Soundbank generation. I am doing some testing and I want to make sure the exact same input file is playing during game playback.

Currently I am using the default conversion settings that sets the Sample Rate 'As Input'. The file I am using is 48k and is 2.7MB before conversion. The output file is still 48k and 1.8MB in size. How do I prevent this compression from happening?

Answer 1

Hi Stephen,

Firstly note that all files will undergo a slight transformation ( even linear PCM .wav files ) into AK's native .wem format ( which is similar to RIFF/WAVE format ).

I'm assuming that your input files are broadcast wave files or have some other wasted data and so when wwise does the conversion the output file is smaller as only the required metadata is included.

If you want to check this yourself; set the file to stream, build the bank and the compare the .wem file with your original file. If you would like me to do the comparision for you let me know.

In conclusion it is very likely that the data saving you are seeing is not the result of lossy data compression, as you fear, but as a result of cleaning up the file.

Hope that helps,

Dan Murray

Comments

Thank you for this answer and that totally makes sense.

I am curious how you would do the comparison. I can't do a diff because the files are binary and I can't import the .wem file into a DAW to do a phase inversion test. What way would you do this comparison?

Also, I want to make sure I am looking at the correct .wem file. When I set it to stream and generated the sound bank there was a numerical file (942515203.wem) in the GeneratedSoundBanks/[Platform]/ folder. Additionally, the number is the ID of the Streamed File in the soundbank xml. It additionally looks like the same size as the "compressed" file that Wwise generates so I am thinking it is the correct one.

Lastly, I did a print from Unity/Wwise out to ProTools over AES/EBU and did a phase inversion comparison with the original file and found that there is some added noise (negligable -125dB) on the recorded file between both Unity on its own and Unity with Wwise integrated. I am wondering if both systems are applying some sort of dither-like noise on the output. I honestly have no idea. I would be happy to share screen shots.

---

Even though its a binary file it still has structure to be compared, I would look at the metadata ( other than the audio samples ) to see what's changed and if that removes a large amount of data or not.

If you want to quickly get the short id for a file shift-right-mouse click on the audio in the actor-mixer hierarchy and copy short id. Alternatively you can perform a query for that short id. It sounds like you have the right one either way.

If your input file is 24 bit dithering or truncation will be applied down to 16 bit depending on the conversion settings you use, the default is to apply dither. Additionally the default is to remove DC bias so be aware of that.