Accuracy of the Change Speed Effect


I was just wondering how accurate the speed effect is. For example, if you were to apply a multiplier of 1.250 (a 25% change), then perform audio editing on your audio (making high level editing faster), and then change the speed back using a multiplier of 0.800, would you get an exact duplicate of the original audio (except for your edits), or would there be some distortion due to multiple speed changes?

The multiplier of 1.250 works out exactly since 1/1.250 is 0.800 and 0.800 fits into the multiplier input box. But if I wanted to process the audio at 135%, then, to get back to the original speed, I would need a multiplier of 0.740740740……. . Since I can only fit 0.741 in the speed multiplier input box, it seems I would not get an exact duplicate in this case – but likely not a noticeable difference?

I know I can use the play at speed option, but I use several macros that are linked to keyboard shortcuts that are linked to my programmable mouse buttons. Those macros end up with the playback speed reverting back to normal speed and so I have an extra step of clicking the play at speed button between each edit to get back to the faster speed. The macros also are tied to the use of the space bar and the “X” key, which basically “reset” the speed back to normal and so again, I need to use that extra step of clicking the play at speed button.

The option of increasing the speed of the whole track, doing my edits, and then decreasing the speed seems like it will work well but I wanted to make sure that no distortion would creep into the audio that way.

Maybe I’m over thinking this, or maybe I’m missing something within Audacity that allows editing at a higher speed with all the same controls as if I was at normal speed.

Any feedback?



That won’t work with conventional sampling because high pitch sounds will slide out of range. It might work with a higher sampling rate such as 96000 instead of 44100.

The sampling rate needs to be at least twice the highest sound frequency. So figure out what the highest pitch sound is going to be at your new speed.

That and there’s sometimes a hardware problem as well. Many soundcards and interfaces cut you off at 20,000Hz (upper limit of human hearing) whether or not the digital system can do it. This regularly kills people trying to record bats or other nature sounds.



Wow - I never would have thought it was that interrelated. I guess I need to look into sampling rates, pitch, and frequency a bit more - not to mention soundcards too.

Thanks for the interesting feedback!



Mostly I’m doing webinar/podcast type editing - so just voice. Below is the spectogram view of an original audio AND the original audio with speed increased by 150% (I likely wouldn’t go that high for editing). The 150% shows a max frequency of just under 16kHz - so does that mean as long as I have a sample rate of at least 32000, I should be okay with my editing approach outlined earlier (normally I use 41000)? And If I were to increase the speed by the 150%, then edit, and then decrease by 0.667 (1/1.500) my audio should be just about the same speed as the original (100.05%; 1.500 * 0.667 = 1.0005) without any significant distortion?

And since I’m under 20kHz at the 150%, I’m hoping my sound card doesn’t have issues.

Thanks again.


The sound card will only have issues if you have to go into and out of the computer.

Pick or make a short demo of voices and music. Send a copy through the process and cut that against a clean version. See if you can tell the difference—of if there is a difference, if it bothers you. Make sure you applied specific tools to go up and different ones to come back down. Don’t just go up and then Edit > UNDO.

Also this is gong to be a problem if your speakers or headphones suck.

Audacity has protections against overloading while a show is loaded, but it has no such protections against frequency or pitch damage. That’s where sampling comes in.


The “Change Speed” effect is very accurate. In most cases it is accurate to the nearest sample. For example. with a sample rate of 44100, the length of each sample is 1/44100 = 0.000022676 seconds (0.02 ms), so Change Speed is likely to be accurate to around +/- 0.02 ms.