[FFmpeg-user] Deterministic results with atadenoise

[Robert Krüger]

Hi,

is anything in atadenoise random? I am trying to get reproducible results
for an automated test setup but just running the same command line twice
produces a files which have differences when comparing them via framemd5. I
thought it might be multithreading but even with threads set to 1 this
differs.

I tried a command line like this:

ffmpeg -i noise_luma20.mov -threads 1 -vf atadenoise -c:v mpeg4 -q:v 1
noise_luma20_denoised_t1_2.mov

Running it twice and then comparing framemd5s of the resulting files, I
find all frames to have different md5 hashes.
Thanks,

Robert