"Peter P." <peterparker@xxxxxxxxxxxx> writes: > Hi Thomas, Måns, list, > > * Dr. Thomas Tensi <t.tensi@xxxxxx> [2023-05-23 17:10]: > [...] >> I understand that this does not directly help Peter with his >> problem. But it means that at least for the effects >> allpass, band, bandpass, bandreject, bass, biquad, compand, >> equalizer, gain, highpass, lowpass, mcompand, overdrive, >> phaser, reverb, treble and tremolo there is a reference >> floating point implementation available as open-source. > Thank you Thomas, this is amazing work! > >> And this implementation reproduces the command-line SoX >> "bit-exactly" (with a residual noise of -150dBFS, as shown >> in the documentation and the test cases), so you can be sure >> that command-line processing and floating point VST >> processing produce almost (!) the same audio. >> >> So if somebody volunteered, a step towards a floating-point >> SoX implementation would be not too complicated... > For the issue I have brought up I would it be sufficient to only > have the software part that reads in the file to not clip the values and > apply the "gain" and/or "normalize" effects before converting to 32-bin > integers for all the other effects. Not super elegant, but possibly > more easier than changing all effects? Unfortunately, that's much more easily said than done. The conversion happens as part of reading the file, before any effects have a chance to touch the data, and different format handlers do it differently. If it were easy, I would have done it already. -- Måns Rullgård _______________________________________________ Sox-users mailing list Sox-users@xxxxxxxxxxxxxxxxxxxxx https://lists.sourceforge.net/lists/listinfo/sox-users
