suppose we have an audio sample of one violin or trumpet note held constant (both in pitch and in amplitude) for 4 seconds.
I would like to extend this duration for example to 60 seconds: the most direct way is to copy-paste the sample (after cutting the ends) 15 times arranging them one after the other. But the effect is that of a turntable needle skipping…
Is there an automatic way to do this? Above all to have continuity in the sound?
Basically you are asking about how to make seamless loops - that’s an art when it comes to real world sounds.
The beginning (the “attack”) of acoustic instruments often differs substantially from the held note, both in amplitude and timbre. It’s usually best to ignore the first part of the note (so that it only plays once), and find a section within the “sustain” part of the sound to loop. For a seamless loop, the beginning and end of the looped sections must match in amplitude, timbre, and “phase” (so that the waveform is continuous from the end of one repeat to the start of the next repeat. Then use the “Repeat” effect to repeat that section (See: https://manual.audacityteam.org/man/repeat.html)
Note that making good loops of real-world sounds is difficult - that why there are so many “sample libraries” sold (pay someone else to make the loops).
Works best if you duplicate the track, apply PaulStretch to the duplicate, then replace the start with the start of the original (and crossfade between the start and the processed sound).
In this “before / after” example, I’ve retained the start and the end of the original, and spliced in a “PaulStretch” stretched middle. PaulStretch does change the character of the sound a bit, but it’s probably the closest to “automatic” extreme stretching. (I also applied some compression after PaulStretch so as to avoid the level modulating too much).
Thanks, but PaulStretch does change the character of the sound a lot…if you try with a cello sound, or a vocal sound, already with a factor of 2 it becomes a synth-like sound, so it’s impractical for the purpose of my question.
Instead, working on the pure original sample, by trying to stitch the seams perfectly like a surgeon, there is continuity. But there is also an unwanted repetition of the micro-variations pattern of the sound, and you can hear it. Perhaps an idea would be (by means of AI-based software) to randomly create slightly different length fragments and then seamlessly join them at the junctions.