GoPro Audio Problem - Any way to restore

browj2 wrote on 1/9/2025, 2:47 PM

Hi,

I have a problem with the audio part of a video file recorded on a GoPro Hero 5. This happens occasionally. All other recordings of the day, about 2 dozen, were fine.

I can hear the sound very faintly, but the waveform shows that there is audio and it looks fairly normal, and the meter shows audio during playback. Here is what I see in VPX16:

Opening the file in Sound Forge Audio Cleaning Lab 4 gives:

so the left channel is fainter than the right channel, but the meter shows sound.

The image below exposes the problem. There is a narrow band between 10 and 15 kHz that seems to be loud, even though I can't hear it well in that frequency range. The spectral display shows that there is other information below 10 kHz of much lower intensity.

Here is the display in Sound Forge Pro 16:

and with the Spectroscope turned on:

where you can see the narrow band at high intensity.

Here is the display in SpectraLayers Pro 11:

If I increase the volume by 30 dB, you can see the lower intensity information:

I made a first run removing the information in the 10-15 kHz band. The result was an improvement, but not satisfactory.

Is there another way to do this to try to get back some kind of reasonable audio?

Here is a link to the file in Magix Hub Explorer. I don't know if the file can be downloaded. First time I've used this. If not, I'll upload it to another share site.

Thanks,

John CB

Comments

browj2 wrote on 1/9/2025, 2:59 PM

As a follow up, I tried using MDynamicEq and got an improvement. I removed the high frequencies and increased the volume on the audio object. So maybe there is some better combination in SpectraLayers.

John CB

SP. wrote on 1/9/2025, 6:51 PM

@browj2 Sounds like most of the full sound spectrum got squished/transposed into the frequencies above 10 kHz. I also tried to remove everything above 10 kHz but the remaining rest sounds thin and has many digital artifacts.

Now it would be very cool if SpectraLayers or other tools like RX could stretch the spectrum image with the audio information between 10 kHz - 15 kHz back out to 25 Hz - 20 kHz. But to my knowledge that's not possible?

Sadly, a tool like Photosounder, which can export and import sound images, is bound to the maximum resolution of image formats. So there are not enough pixels to get a useful result after stretching and reimporting the image.

browj2 wrote on 1/10/2025, 6:59 AM

@SP.

Hi,

Thanks for taking a look. It's not all that important that I get good sound as I'll probably just use music over snippets of the video, but some would be good. This was one of the few sequences that I shot with only one camera. In any event, I had the wrong audio setting on my DSLR with a mic attached or something as there was a lot of clipping in most shots. The sound in the room was far too loud for most ears and my mics.

I would like to know what causes this problem occasionally. I don't think that I had the wireless mic connected to the GoPro for this sequence, but I may have. I'll have to do some testing.

I'll spend some time this morning going through the SpectraLayers manual and trying to bring down those high frequencies and raise the lower ones. It's time to get educated on the uses of this program.

Thanks,

John CB

johnebaker wrote on 1/10/2025, 7:22 AM

Hi John

Could this be intermittent external interference while recording?

Were you using the GoPro built in mic, or an external mic connected to it?

I no longer have the video file, deleted as soon as finished testing, however IIRC for the first 'sound issue', I believe a guy walked in front of you holding what looked like a mobile and was recording, the second time there was nothing obvious other then you being extremely close to a camera on a tripod.

John EB

browj2 wrote on 1/10/2025, 8:24 AM

@johnebaker

Hi John EB,

I was able to boost the audio as I showed, quite well, in fact.

If I had the wireless mic on, it would have been on the groom (purple shirt). I don't hear him speaking at any point, so the wireless mic was probably not connected.

I was mistaken, I did have the other camera on (my camera on the tripod) - the crowd watching at the left. The sound is extremely distorted. Maybe I can also do something with that in SpectraLayers. I have to check what went wrong there. I had a mic (Opteka VM-100 condenser shotgun) on the camera (Nikon P950). I just looked and the pad is set to 0 dB. I probably should have had it a -10 or -20 dB. Auto sensitivity in the camera was on.

Here are the waveforms on the timeline.

The DJ was walking around with a wireless mic and he was filming occasionally with a cell phone with a light on it.

As for the video part, the lighting was poor. I had one light, on the DSLR that I pointed where needed. In the current clip, it's on the action being filmed with the GoPro, so the part with the DSLR is very dark. Bringing it up leads to a lot of video noise. I am going to need some help from Neat Video and this current bug is very annoying. I hope it's sorted out soon.

For anyone reading this, another audio problem I had was with a speech by a woman using the same mic as the DJ. There was a fair amount of reverb and some background noise. I used Accusonus ERA6 ReverbRemoverPro and NoiseRemoverPro - lightly. Too much introduced artefacts. Being able to remove or reduce reverb is quite an advancement. I am going to try DeReverb and DeNoise in SpectraLayers to see if I can do better.

Thanks,

John CB

browj2 wrote on 1/10/2025, 1:38 PM

@SP. @johnebaker

Hi,

Here is my effort after SpectraLayers 11 then SFACL4. I tried fixing the SPL11 file in SFP and it was a mess. Sorry, but I can't figure out that program, even with the manual.

In SpectraLayers 11, I removed the strip of high frequency sound, put it on another layer and muted it. Then, I unmixed voice, then in the noise layer, I selected a fairly dead spot in the recording and used it as a noise print for the Denoiser which I then applied, gently. I had already boosted the volume, but boosted the volume of the Voice a bit more, decreased the volume of the Noise layer a bit and exported. I got a file for SFP. I opened that in SFACL4, used Melda MDynamicEQ to reduce some high end

In the image below - original on track 2, result from SPL11/SFACL4 on track 5. Somewhat astounding.

Here is the link to the folder with the wave file on Magix Hub. Just found that I can't upload a wave file to the forum directly.

Please let me know if you are able to download the file.

John CB

johnebaker wrote on 1/10/2025, 1:45 PM

@browj2

Hi John

. . . . The DJ was walking around with a wireless mic . . . another audio problem I had was with a speech by a woman using the same mic as the DJ . . . .

That could be the culprit, especially if the mic was a 'long range' ie higher power one.

John EB

rraud wrote on 1/11/2025, 11:08 AM

Now it would be very cool if SpectraLayers or other tools like RX could stretch the spectrum image with the audio information between 10 kHz - 15 kHz back out to 25 Hz - 20 kHz

It may be possible using SLP's clone tool. I use it (clone) occasionally to fix a frequency specific glitches and such. I do know how it would work in this particular case though.
An alternative to using clone tool is copy/ pasting the high frequency content down to the lower spectrum

SP. wrote on 1/11/2025, 1:08 PM

@rraud Here are two spectrum graphics (orignal and edited), exported by Photosounder, where I scaled the high frequencies in Gimp down to the bottom of the sprectrum. This more like a non-uniform pitchshifting. The sounds at 10 kHz move down to under 100 Hz while the sounds at 15 kHz basically stays at 15 kHz.

But in the end this doesn't lead to any listenable result when coverted back to an audio file. Maybe the audio wasn't squished but is simply overblown in the high frequencies because of the wireless interference as @johnebaker speculated.

browj2 wrote on 1/11/2025, 1:36 PM

@johnebaker

Hi John EB,

Not to confuse the two issues. The recording of the woman with the mic was fine, except for the effect, reverb, that the DJ likely had on the mixer. See ** below. I just used the DeReverb a bit and also got rid of some background noise. I must try this recording at some point in SL just to see if I can do better. The bad recording is the one that I posted. I don't think that it was due to interference with the DJ as I have had this happen a couple of times in a completely different setting.

@rraud

I tried the clone tool and got nothing. I watched the Steinberg SL cloning video tutorial and it seems that cloning just takes the source and reproduces it at the same frequency, overwriting whatever is there. Please let me know if you find a way to do that can clone at a lower frequency.

I tried selecting a rectangle of the band between 11 and 15 kHz, then used Transform, and gave it a lower frequency, about -9 kHz. In the image, the first part was done twice, the second time once. The 2 cyan rectangles were the source areas. You can see that I have a wider frequency range than the original. The audio sounds like shakers or a tambourine. No other useful information.

If you listen to the result that I posted, you hear that my method of removing that high frequency band, separating out Speech from Noise plus some noise reduction gave me a very good result, relative to what I started with.

I tried the audio from the other camera and my result was fine for part, but when the music was playing, it was clipping too much making it unusable. There is a pulsation in the music itself that becomes exaggerated in the recording.

In any event, all I needed is a few snippets of the DJ, and some background crowd noise. SL gives me this. Then, I can mix in the original music. The result will be an intro by the DJ, the background music with ambient crowd noise not too loud, and edited shortened clips of the action cut to the beat. Result - probably 30 seconds or less instead of several minutes.

This has been a very interesting exercise.

** I can see some other things that can be done, like removing other voices that shouldn't be there or that I don't want there. In the speech by the woman, there is some background talking that I would like to try to remove. Once could spend a lot of time on just editing audio like this.

@SP.

As I indicated above, I didn't find anything useful in that frequency band either when I lowered the frequency.

Thanks,

John CB