Reply #45: UPDATE: Can now NOT say whether it was doctored (my brain hurts) [View All]

[anigbrowl]

OK, there are differences between your version and the original versions off YouTube (which I just now recorded using your links). As your thread had been locked by the time I got hold of the wav file, I didn't see the bit about the various changes you had made, which could have added extra spectral info to the file, depending on what they were.

So, I recorded both YouTube clips (the disputed one and the longer extract from the original, which I will henceforth refer to as A and B) and they do NOT show spectral evidence of doctoring - either of them. That is to say, the audio characteristics of both appear to be identical. This does NOT mean they are identical, but that there is no obvious evidence of tampering.

I DID find marginal differences between the two. These could easily have arisen from the fact of the two different clips being encoded on two different computers - unless they were both using the exact same software operating on the exact same clip, you would expect this.

B was somewhat louder than A. I obtained the peak amplitude level of B and normalized A to that amount (that is, i scaled the maximum audio peak of Be to be the same as the maximum audio peak of A). This is just a volume change. Having done so, I noticed little difference. Viewed side by side, peak values of both waveforms match well down to the sample level. Lower amplitude waveforms match approximately in shape but not in amplitude. I tentatively ascribe this to differences in the encoding/compression algorithm, as above.

As a further text, having aligned and locked the waveforms side-by-side so the peaks matched, I inverted A (which I will refer to as A-), then mixed A- with B to see if they cancel (effectively subtracting A from B to produce a new waveform, C). They do not, but this is reproducible with differing codecs as described above. As a test, I inverted C to produce C_, and mixed A- C- B. As expected, this resulted in silence. (the point of this last test was to make sure it wasn't some deficiency in my software - I know it's sample accurate, but I just wanted YOU to know that I checked).

OK, the big question you are asking is "what about the phrase which sounds like 'worthless white nigger'" and the answer is that it DOES appear to present on both waveforms, A and B. That is, it's audible on both the controversial clip and on the longer, uninterrupted segment, both of which I recorded from YouTube.

In short, I was WRONG to say with confidence that YouTube clip had been doctored, as my observations were based on a clip supplied by Pawel (which he has told us he did enhance for clarity purposes) as opposed to an untouched recording from YT. I can NOT say with confidence that the material on YouTube was doctored in terms of having extra audio dubbed on top.

You can argue among yourselves whether pausing and adding subtitling constitutes doctoring. I call that analysis. All I can tell you is that my statement about dubbing having occured based on spectral signatures is NOT correct, as I had downloaded a copy which had been filtered by Pawel K (in good faith, I hasten to add).


I would have to check the DVD to be sure of anything beyond that. It is possible, for example, that the clip A on YouTube was derived from clip B and re-encoded (which would introduce the encoding differences I referred to above). IF clip B (the longer, unedited, unsubtitled extract from the film on YouTube) is an accurate copy, then it would APPEAR that the words in question are present on the soundtrack after all.