Quais são algumas técnicas eficazes para reduzir o ruído e os sons de fundo no áudio de voz?

The point about reviewing voice over audio in different speakers and headphones is often overlooked but so essential. Different devices can highlight different aspects of the audio, and what sounds clear on one might reveal issues on another. It's always a good practice to test audio on multiple devices to ensure consistency and quality across the board.

Determining the source of the noise is crucial, and obtaining a sample of that noise is also important, as it constitutes the next step in the noise removal process.

Important to mention is reverb as source of "noise". The amount and "quality" of the reverb can vary greatly. It depends on the location, microphone type, polar pattern and of course, mic distance and placement relative to the voice talent. Then you should consider the purpose of the recording since some productions might tolerate more reverb than others. Game VO for example should be recorded as dry as possible since no room information is wanted but rather added later. Podcasts can tolerate a bit more room information since listeners assume people are in a room and it's part of the ambience.

Izotope RX is the leading software used in voiceover settings to remove audio issues. It takes some time and patience to learn but also has some out of the box features that make de-essing, de-clicking and de-noising very quick and efficient.

es ist oft nicht so einfach gerade bei mehrere Störfaktoren die Quelle zu identifizieren. Wenn man sich nicht bei der Aufnahme vermeiden kann, hilft ein Spectrogramm das Identifizieren von Brummen etwa sind Linien und ihre Frequenz-Vielfache. Tools wie De-Construct helfen das Signal in ihre Tonalen und Rauscheanteile zu zerlegen. Das kann noch einmal helfen Störquellen zu identifizieren

Keep it sounding natural. Make sure you don’t overdo it with the noise reduction. Revisit your toolset often. The quality and speed of noise reduction software is getting better at an exponential pace. Be sure to try new things to achieve the right sound for the voice you are working with.

Please exercise caution when removing noise. It could lead to sound distortion. Benefits and drawbacks must be assessed. Sometimes, leaving some noise untouched maintains the natural sound, while removing all noise can result in distortion.

Modern noise reduction tools work better as they use neural network engines behind the scenes. However, not all of them work the same. Because they are trained in different data sets or different types of noises. A few examples are Waves Clarity VX, Supertone Clear, iZotope RX, etc. There are more, but I named a few. Traditional noise reduction in Audacity or Adobe Audition does not have massive training data sets. You should try any of the modern noise reduction tools. Though they are paid plugins, each one of them offers a free trial. If a software works best for your voice, that is the software for you. However, you have to keep an eye as things are upgraded with more training datasets and more advanced technologies.

The challenge often can be to draw the line between what is unwanted noise and what is present but tolerable. It can be based on how loud and distracting the sound is, how much postprocessing will still be applied later in the chain that might worsten or mask unwanted sound or in what context the voice will be used.

Use tools in phases, like a photo editor grading and colorizing a raw picture file. Split sections of audio with different noise profiles before working on them individually. Start with the things that are the easiest, clicks and pops. Then move to de-essing and breath control and perhaps some very light compression or volume automation to smooth out the section you're working on. Then do de-hum or de-noise after, because the frequencies removed in the previous processes are not going to hit our noise removal algorithm, which makes the noise easier to pull out. Manually tweak your settings so you don't run it at max value at first. Do it in layers, you won't lose anything from running it through again, and if you do, go back.

Please be careful when using those tools. Generally, noise removal won't be perfectly accurate compared to manual remove. It's often better to remove subtle noise manually.

Using a De-esser tool requires knowledge and understanding because, unfortunately, it can sometimes remove a significant portion of high frequencies, affecting sound quality and clarity. In my opinion, evaluating the need for this tool is crucial; often, manually removing disturbing sounds, especially in letters like 'S,' is more than sufficient. For issues with letters like 'P,' manual intervention works better by selectively removing the problematic frequency part. This problem can be avoided by recording properly and maintaining an appropriate distance from the microphone during recording.

These tools can be really helpful, but the first step must be to avoid plosives and sibilance at the point of recording. De-esser and de-plosive plug-ins can really affect the natural sound of a recording and have their limitations, so it's best to focus on getting the microphone placement and technique right before recording.

Always ensure your sound falls within the range of -6 to -3 decibels. These values might vary slightly at times, but the sound shouldn't be consistently loud or quiet. This is not just about noise but also about creating an acceptable listening experience for podcast, radio, and voice-over content.

This step is important even though it's not directly related to reducing noise. Exporting audio files according to the standards of the platforms where the content will be published is essential. Proper and organized naming is also highly important.

Your starting audio must still be pretty good. As others have mentioned you can overdue corrective measures and it will damage the audio further. Good clean audio from a quality mic and interface in an acoustic-treated space should need little to no post-production enhancement.

It’s important to keep an eye on the signal chain from the source. Is the microphone directional? If you’re recording and your microphone is set to fig8, you’re definitely going to get a lot more pickup from behind the microphone. Is the room properly isolated with acoustic foam? Is the shock mount correctly isolating the microphone? If you find a lot of low end interference this may be something to look at. If you’re on location use a shotgun mic backed up with a lav to be sure. What I’m trying to say is that noise reduction isn’t something that you should only start in post. You get out what you put in!

It's also vital to ensure you're the right distance from your microphone, and that your microphone is placed correctly during recording. Some microphones record sound from a particular direction, and some have settings that enable the microphone to record sound coming from every direction. Spend some time understanding your microphone as it may have tools built in to limit background noise. Similarly, sit the right distance from the microphone. If you're too far away, there's likely going to be more noise.

Throughout my career in optimising voice recording, the recent challenging times have underscored the value of tools like RX. I've transformed the audio quality of countless videos for clients using this software. To this day, the Dialog Isolate module feels like pure witchcraft to me.