Film to PAL Speedup Audio pitch/tempo quality

[mp3dom]

Hi to all!
I need to speedup an audio at about 4% for Film to PAL video transfer. My first idea was to keep the original pitch so I have applied a timestretch but the end result was not so good because some artefacts appears (mainly echoes on the dialogues part). So I have changed my approach and go with the "old way" and I've simply speeded up the audio by 4% obtaining a frequency of 50 KHz, then resampled at 48 KHz. Since the audio is mainly music I have decided anyway to restore the original pitch, so I've decreased the pitch by ~0.72 semitones. This audio seems to not have echoes or strange artefacts, having tempo/pitch the same as the timestretched version.
I know that timestretch is "lossy" because introduces artefacts, so I would like to know, quality-wise, where a quality loss occour in the "old way" method.
I assume that speeding up the audio to 50 KHz is lossless (excluding the pitch/tempo change, I think no artefacts are introduced) right?.
Resample introduces artefacts but a good resampler could limit the artefacts to about zero so I think is a good compromise.
What I don't know is if changing the pitch (keeping the tempo original) introduces heavy artefacts or if it's near to lossless.
Thanks!

[Inventive Software]

With AviSynth, have a script that loads your video, and the audio, and then add "AssumeFPS(25,sync_audio=true)".

This shrinks the audio duration about 4% and resamples as a result, however the audio will sound perfect if you play it through. Assuming your target format is DVD, after the AssumeFPS command, add "SSRC(48000)", and you should have perfect audio.

Hope this helps. :) I do the same sort of thing, except in reverse, to get PAL to 24 FPS.

[mp3dom]

Thanks for the response.
Yes it's similar to what I do... after speedup (without the SSRC) the frequency is 50 KHz. Then with SSRC, resample to 48 KHz. But the result sound different (about a semitone) in pitch and this is quite annoying if there is a lot of music in the film so my question involve mainly the pitch shift process (changing the pitch after the resample). Is it a process that can introduce artifacts? Is there a way to obtain the max about pitch shifting quality? Dunno why but the timestretched file (based on MPEX at max quality) sound worst than the file manually resampled and manually pitch shifted (which has less echoes, but still present) here my question about getting the max quality from pitch shfting.
Thanks! :)

[Inventive Software]

When you pitch shift it will always sound different to the source. SSRC is the best re-sampler I know of, so that's why I suggested it.

[madshi]

Most studios do not pitch correct their PAL sped up audio tracks because pitch correction adds ugly artifacts on its own. If you insist on doing pitch correction, you might want to check out TimeFactory, which I've been told has the best pitch correction algorithm currently available.

[nautilus7]

When i convert ac3 audio from 25 fps to 23,976 fps with besweet or mkv toolnix (during muxing process) do i re-encode it also?

[madshi]

When i convert ac3 audio from 25 fps to 23,976 fps with besweet or mkv toolnix (during muxing process) do i re-encode it also?
BeSweet does reencode. mkvtoolnix I believe does not reencode but it stores the original audio track with additional "stretch" information. Of course that means that sample rate must be changed at runtime then.