Automatic stitching of Cinema DTS files

[FoLLgoTT]

I created a Python script with AI which stitches the decoded parts of a Cinema DTS CD to one large multi channel audio file.

Features:

• Detecting the signal tone and removing unneeded silence
  
• Detecting overlap by correlation (with detection of inverted parts)
  
• Short cross fade for reducing audible errors
  
• Merging of the files
  
• Extraction of LFE including clipping protection and custom LFE gain (optional)
  
• Slowdown to 23.976 FPS and resample to 48 kHz (optional)
  

It works good so far, but there are several CDs which have serious mastering errors. I even found one without a signal tone and one with a DC offset (!). So don't expect that all CDs can be merged without problems. There are some parameters to play with to accept lower correlation or remove longer silence before the signal tone.


CinemaDTSFileProcessor

[stwd4nder2]

Interesting. I always kind of assumed that automating this wouldn't be possible with how much the silence between reels can vary, but I never really looked into this much.

Is it possible to do a slowdown without resampling (if it's even worth it/a good idea)? Also I wonder if it's possible to integrate AUD to wav conversion and just have a single script take care of everything.

[FoLLgoTT]

Interesting. I always kind of assumed that automating this wouldn't be possible with how much the silence between reels can vary, but I never really looked into this much.

Yes, the silence varies. The signal tone varies in level and channel (!) and even the silence before the signal tone varies. Sometimes there is even a signal tone in the beginning of a file. But I could solve these problems. For most CDs it works pretty well. And from my experience newer movies work better.

Is it possible to do a slowdown without resampling (if it's even worth it/a good idea)?

I don't think so. The sample rate must stay at fixed values (e.g., 44.1 kHz or 48 kHz). You can only speedup the FPS of the video losslessly with FFMPEG.

Also I wonder if it's possible to integrate AUD to wav conversion and just have a single script take care of everything.

I will evaluate this.

[CSchmidlapp]

How easy do you think it would be to introduce a sync feature where you feed it a reference audio track to sync to?

[FoLLgoTT]

How easy do you think it would be to introduce a sync feature where you feed it a reference audio track to sync to?

I already implemented a simple version of "FindDelay" using correlation. But currently it is not reliable. Especially if the reference track differs too much (e.g., by dynamic range compression).