I am new here and, quite apart from that, I’m an oldie and not au fait with music jargon. I work with Win 10 Pro and have downloaded Audacity 3.1.0.
My first problem is that I still have to earn my living, secondly, that I do this in part by transcription. Sadly, the people who record interviews, group discussions, presentations, etc. are not very good at it. For instance, with most interviews, the interviewer is sitting bang up against the microphone whereas the interviewee sounds as if he is on the moon without a microphone at all. Other problems are background noise and reverb.
I have been trying to solve these problems with Audacity but it is really difficult if you are not familiar with the terminology. Is there anywhere I can gen up on background noise, reverb, etc. so I can get the settings right from the start. As it is I spend ages trying things out when I should be working on the actual assignment. What would be helpful is either a website with explanations of the various settings, e.g. Damping, Tone low/high, wet/dry gain, stereo width. Alternatively, do you have a selection of pre-sets somewhere which I haven’t discovered yet?
Advance thanks for any suggestions or links.
Thanks to anyone taking the time to read this, and sunny greetings from Berlin.
it is really difficult if you are not familiar with the terminology.
It’s difficult if you are. The kiss of death forum posting is one that starts: “Help Me Clean Up…” As a fuzzy rule, if you can’t understand both sides of a conversation at the start, none of our tools are going to help you.
interviewer is sitting bang up against the microphone whereas the interviewee sounds as if he is on the moon without a microphone at all.
Right. That’s the Skype/Zoom/Meetings user who thinks they can record an interview at home. Because of the way those services work, good home recording is difficult/impossible.
background noise, reverb, etc.
The short version is we can’t do anything about reverberation or echos. That’s the performer’s voice arriving at the microphone more than once having bounced from the walls and ceiling of their studio, or worse, internet echos. I recently conducted an internet service call where my voice came back to me loudly two seconds late. That’s profoundly annoying/confusing.
I spend ages trying things out when I should be working on the actual assignment.
Your ‘job’ is straightening out and interpreting the damage. You can get a machine to transcribe clean, clear voices. I don’t know of any shortcuts, classes, or software manuals.
Thanks for your reply and thank you also for being upfront about my chances - or rather the absence of - of solving the problem with Audacity.
As to your remark, “You can get a machine to transcribe clean, clear voices.” You can? That’s the first I’ve heard of it. Could you possibly name a couple? I hope these are not software solutions that cost an arm and a leg. That would really ruin my day - week even!
It’s extremely difficult to fix problems such as background noise / echoes / distant voices / distortion even for experts that are familiar with all of the terminology.
The first, and arguably the most important step in producing clear, high quality audio recordings, is to start with a good, clear recording. You can polish a gem, but to mix metaphors: “garbage in / garbage our”.
Important points when recording:
Record in a quiet room that is not echoey.
Ensure that you are close enough to the microphone (the best distance depends on the mic, but for typical voice recording mics, around 10 to 20 cm is in the right ballpark)
Avoid blowing on the mic
Set the recording level carefully so that you have a reasonably strong signal but (with digital recording) never go into the red.
Use reasonable quality equipment.
Note that the recording environment (a quiet, echo free room) is at the top of the list.
My advice would be: get familiar with the noise reduction tool.
If the background noise is constant and uniform (humming etc…) → try to remove it with the noise reduction tool.
If the background noise is anything else like people talking, walking around etc → don’t waste your time, you’re not going to get good results without affecting the speech parts.
When it comes to reverb, you can try and fix some of it with the noise reduction tool.
Select the “reverb tail” (the part of the audio at the end of a word / phrase that consists only of reverb) as your noise profile and reduce it. you can reduce the reverb a little bit, but don’t expect perfection. You can’t polish a turd.
With the people too far away from their mic, you can try a lowcut filter to reduce some bass frequencies and boost high frequencies above approximately 1-2kHz so the speech becomes a little more intelligible. It depends on your audio, you just need to play around with the EQ to find what works best. But again, don’t waste your time. A bad recording is a bad recording is a bad recording. Frustrating, but not much you can do about it.
Thank you for your reply which comes at exactly the right moment.
Just to be clear, I don’t do interviews myself, I transcribe and translate them. Bad quality recordings make this type of work very difficult and increase the time you need to spend on an assignment.
A colleague of mine who is experiencing the same problem has expressed a need to educate clients. I shall, if I may, send her your list - maybe she can use it in a mail to her customers and make our lives easier.
It’s not uncommon for professional transcription services to charge extra if the recording is so bad that it is difficult to hear what is being said. That’s a strong “encouragement” for their clients to take a bit more care with their recordings
Freelancers - and that’s what we are - have very little leverage. We have to take what comes and grin and bear it. We can try and “educate” our clients to provide better recordings but when it comes to the mentioned “encouragement” I don’t know that that would have much effect and we have to be mindful about keeping our clients. It’s a tough old world.
Anyway, a big THANK YOU to everyone for your input. I have to get back to work as someone has announced a (hopefully good) recording for later this afternoon.
All the best to all of you and cheers from Berlin.
What I always say is, “The brain is usually the best filter”. In most cases, the best & fastest solution is to just stop and re-listen a few times. And I assume a bad recording takes more than twice as long to transcribe. You may be able to do some quick filtering & processing but that often takes trial-and-error.
The situation is a little different if your goal is to improve the recording (for future listening) but if the recording is very-bad it’s still often helpful to have a transcript (or subtitles with video) to go-along with the “improved” recording.
Could you possibly name a couple?
I’ve never tried it but [u]Dragon[/u] has been around a long time. (I don’t know if there’s a German version.) And, there may be some online tools. Siri and Alexa are very good at voice recognition (running on powerful servers on the Internet) but I’m not sure they (or something similar) can be used to transcribe.
You’d still have to identify/separate the speakers and probably do a fair amount of editing so I’m not sure if it would save time, and I assume you’re already pretty-fast at the keyboard… “Theoretically” automated software should be able to analyze a digital recording faster than real-time without actually playing the audio but I don’t know if Dragon does that.
It’s not uncommon for professional transcription services to charge extra if the recording is so bad that it is difficult to hear what is being said.
That’s what I was thinking. Or maybe raise you standard price a bit and give a “professional discount” for “professional recordings”. But, I don’t know anything abut this market. I’d guess it takes around 2 hours to transcribe a good 1-hour recording? More? And I guess a bad recording takes twice as long (or more) than a good recording.
