question about audio leveling

[schmosef]

Hi Guys,

I have a customer who has a PowerPoint presentation that links to a handful of mpgs that he made via his Sony DVCamera.

The mpgs are all interviews of people and he didn't use a remote mic so although the interviewer, holding the camera, sounds fine, the interviewee tends to sound very faint--understandable, but faint.

He's asked me to find a way to fix the audio levels in his mpgs and ideally, create a simply procedure that he can use to fix new files in the future.

It's been a while since I've done anything like this. I've used SoundForge in the past to level wav files on custom CD compilations and I'm assuming I can find a program to strip the audio in the mpg files into wavs, fix them, and then reintegrate them. But I was thinking that there is probably a better technique available.

I'm hoping that some of you guys here could point me in the right direction.

In case it matters, I believe he captured the mpgs to pc via Microsoft Movie maker in some cases and Ulead VideoStudio in others. He owns VideoStudio but is just a novice.

BTW, it's ok if he needs to buy a certain software product to do this to get the best results. He's not expecting this to necessarily be for free.

[Brian Ellis]

IMHO, there is little chance of a really satisfactory solution. At the best, it will be a palliative.

Most cameras have AGC so that when the interviewer is speaking, the sound will be great and the background noise will be largely suppressed. When the interviewee is speaking, it is probable that he won't even be peaking to max, but the gain will be at max, so all the background noise is possibly only 10 dB or so below the voice.

It is possible, of course, to jack the level of the voice up, but the signal-to-noise ratio will not be better, nor will the added acoustics (if recorded indoors).

If the noise level were, say, 40 or 50 dB below peak, a noise suppression filter would work just fine. If you tried it at anything under 20 dB, you would not only get 'orrible distortion (akin to crossover distortion), but you could lose whole words, where the voice dropped.

I think that the only thing to do is to work on each segment of interviewee's speech manually, practically word by word, tweaking it to obtain the best compromise. A very long, painstaking process, I'm afraid. You will require any good audio editor which shows the waveform accurately and with a set of good filters.

Sorry if I can't give you the easy answer.

[SpiralDragon]

ay.... sory to say that its a painstaking procedure with a not so clean outcome..... what brian said is true... pluse you have to take into consideration the delay in in distance effect that sort of creats a reverb effect due to the acustics of the room/venue... that cant be removed.... i would however sugest trying the foollowing... boost the signal where the people talking are not heard or are far away... cut off the high frequances and lift the midrange ... experiment withe the low frequancy till you get a 'satisfactory' result. duplicate the chanel pan each one exream left and extream right withone of them slightly delayed.. like max about 5ms... this gives an intresting stereo fx and brings out the voice slightly by sort of dubling it....

[schmosef]

ok, thanks for the advice guys.

[schmosef]

Thanks for the heads up about the background noise.

By some turn of luck there was enough range between the interviewees and the background so that I was able to amplify the volume and then apply a noise filter that did a reasonable job of cutting out the hiss. Also, they cut the interviewers’ questions out of the video, I guess to create a more spontaneous feel to the interviewees’ comments; so I didn't have to worry about distorting the interviewers’ voices.

I used Windows Movie Maker to extract the audio and after trying a few editing programs listed on Tucows, I found a program that I already owned, Nero’s Wave Editor—bundled with Nero 6, which I used for the amplifying and noise reduction. I was able to raise the volume between 18 and 30 dB, depending on the original source, without encountering too much distortion. The noise reduction was great. The interviewees ended up sounding just a little metallic for the first few seconds on a couple of the videos but the effect doesn’t last long and it’s not distracting so it’s no bother.

I reintegrated the audio by creating a project in Windows Movie Maker that combined the video in the original wmv files (was told they’d be mpgs but was given bad info) with the audio in the wavs I created. I even added an intro circle transition and an exit fade for some additional flair.

I’m very happy with the results. I know my customer will be as well. The PowerPoint presentation was developed by a charity to present during funding drives. My time was donated for this.

My appreciation guys.