Audacity Forum

Why can't Audacity and other audio applications make accurate speed changes? You can test it for yourself make a 2 min audio file then increase the speed to 200%. Easy right? 1 minute. Not according to Audacity, 43 seconds. Huh?

I need to do very small speed change to sync a portable digital recorder with my video camera. Doing a Tempo change seems better but there is something wrong with my math and the way Audacity measures tempo. I was expecting 100% should be normal tempo but instead Audacity uses 0%. If someone could check my math the audio recorder is 40 mila seconds an hour slower than the camera. So the audio recorder is 3,600.40 seconds long and the camera is 3,600. How would one figure out the correct percent to change in Audacity's Tempo Dialog box?

wickedpete wrote:Why can't Audacity and other audio applications make accurate speed changes? You can test it for yourself make a 2 min audio file then increase the speed to 200%. Easy right? 1 minute. Not according to Audacity, 43 seconds. Huh?

The % figure in the "Change Speed" effect refers to the "change" in speed. Thus to speed up from 2 minutes to 1 minute you are making the speed 100% faster.
I agree that this is a bit confusing, but it has been designed with turntables in mind and in this context increasing the RPM of a turntable by a particular percentage makes sense.

I believe their is a plug-in that does speed change but has a user input that has the speed change as a percentage of the original speed, thus normal speed is 100%, double speed is 200% and so on. You may be able to find the plug-in with Google.

wickedpete wrote:I need to do very small speed change to sync a portable digital recorder with my video camera.

You will probably have trouble getting the accuracy that you require.
You will definitely not get sufficient accuracy using the "Change Tempo" effect - use "Change Speed" instead.

The calculation is:
(C/N) x 100
Where
C=Change in length (original length - new length)
N=new length

wickedpete wrote:If someone could check my math the audio recorder is 40 mila seconds an hour slower than the camera.

It is very unlikely that either the audio recorder or the camera run to an accuracy of 40 ms/h
Also, if the sound and images are only 40ms out of sync it will probably be barely noticeable.
The best approach will probably be to run the audio and video together and make a note of when they appear to go out of sync, then make a small adjustment to a portion before that point to drag them back into sync. Repeat as necessary. (a speed change of 0.01% applied to a 10 second selection will change the length by approximately 10ms)

stevethefiddle wrote: The % figure in the "Change Speed" effect refers to the "change" in speed. Thus to speed up from 2 minutes to 1 minute you are making the speed 100% faster.
I agree that this is a bit confusing, but it has been designed with turntables in mind and in this context increasing the RPM of a turntable by a particular percentage makes sense.

I believe their is a plug-in that does speed change but has a user input that has the speed change as a percentage of the original speed, thus normal speed is 100%, double speed is 200% and so on. You may be able to find the plug-in with Google.

I tried it out and yes you're right now speed works but I must say from a math stand point its confusing.

You will probably have trouble getting the accuracy that you require.
You will definitely not get sufficient accuracy using the "Change Tempo" effect - use "Change Speed" instead.

The calculation is:
(C/N) x 100
Where
C=Change in length (original length - new length)
N=new length

Thx, right again I'll have to write it down so I don't forget. I made the mistake of dividing the two times into each other. Tempo change was accurate enough for my purposes.

It is very unlikely that either the audio recorder or the camera run to an accuracy of 40 ms/h
Also, if the sound and images are only 40ms out of sync it will probably be barely noticeable.

On this account you are wrong. In video, you work in frames. There are 29.9 frames a second thus 1 frame = 3.34 mila seconds. Anything greater than 2 frames you hear an echo. If you don't believe me try offsetting the R and L channels of a song by 40ms and give it a listen with headphones.

Anyways thx for helping me figure things out. One thing I give Audacity props for is it works and does it quickly. Many paid for programs wouldn't work because they would round off the decimals. Adobe Soundbooth did it but took 2.5 hours to change the tempo of a 1hr of audio that Audacity could do 8 minutes! Yikes maybe it does a better job but that's a lot of time! And Apple pro sound editor doesn't even allow you to change the speed let alone tempo by a percent.

wickedpete wrote:There are 29.9 frames a second thus 1 frame = 3.34 mila seconds.

33.4 milliseconds.

If you have 2 identical audio tracks playing simultaneously and you then delay one of the tracks by as little as 10 milliseconds you will hear a very noticeable "phasing" effect. It becomes a noticeable "echo" when you get to about 50 milliseconds. This is assuming that you can hear both the original and the delayed signals simultaneously.

In professional audio (and video) it is common to synchronise devices using a "word clock". The word clock is designed to provide a common, accurate and stable timing signal that is used by all the devices to keep them synchronised. The typical accuracy of such a device is around 1ppm (0.0001%). Domestic grade devices have internal clock signals that are usually well below this degree of accuracy, and even more importantly, they drift.

You are talking about an accuracy of 40 milliseconds per hour, which is 0.04 seconds per 3600 seconds, which equates to about 0.001%. I think you will be lucky if both your audio recorder and camera maintain this degree of accuracy consistently. It is not uncommon for on-board sound cards on computers to drift by several seconds over a period of an hour (though a reasonable quality sound card or stand alone digital recorder should be a lot better than that). So all I was saying was that the difference in speed between the audio recorder and the camera may not be entirely consistent over the entire duration, though most probably sufficiently consistent as to not cause a problem.