6 neural networks to improve sound quality

1. Auphonic

• Platforms: web.

An online service specializing in post-production tools. Auphonic can be used while editing content for podcasts and videos. The set of available algorithms includes not only the removal of extraneous noise, but also volume normalization and automatic balancing between different sound sources. For example, you can bring the recording of speech to a level that is convenient for perception compared to background music.

Auphonic works with single files, although it is also capable of multi-track processing. At the same time, for each track, you can specify a spatial characteristic - whether the sound should be in the foreground, if it is a voice, or should it remain in the background.

The tool also supports adding metadata to the output file. In the details of the recording, you can specify the authors, participants in the recording, album name, publisher, license type, release year, genre, tags and comments.

In the processing settings, you can connect an external speech detection service. This is useful in case you need to move the recording of voices to text format automatically. The parameters of the same type can be saved as a template for later work.

Auphonic is free to use, but the total duration of processed audio tracks will be limited to 2 hours per month. For professional work, you will have to connect a subscription costing from $ 11 per month.

Try Auphonic →

2. Audi Studio

• Platforms: web.

Audo Studio is similar to the previous service, but in this case, the focus is on filtering the voice recording from extraneous noise. In addition to removing annoying backgrounds, the tool can neutralize echoes and reverberations in a room, as well as even out the volume level. There is one slider to adjust the processing intensity.

The finished result can be listened to and quickly compared with the original version. The processed file can be easily converted to all popular audio formats: WAVE, AIF, OGG Vorbis, AAC, MP3, M4A.

For free, Audo Studio allows you to process up to 20 minutes of recordings per month. For $12, the limit expands to 10 hours per month.

Try Audo Studio →

3. Krisp

• Platforms: Windows, macOS.

Krisp acts as an additional layer between incoming microphone audio and audio applications. online conferences. The program neutralizes the echo, filters the noise of cars, the screams of children, the background speech of people, the hum of fans, the barking of dogs and other sounds that may distract from business calls.

The system automatically detects which application is currently being used for the call and sends the purified audio to it. In this case, you can filter out the signal not only from your side, but also incoming noise from other participants in the conversation. This additional feature, of course, only works for your peace of mind and cannot interfere with the overall audio of the conference.

The application can be used for free for 60 minutes a day, and for the removal of the restriction, the developers offer to subscribe for a price of $ 96 per year.

Try Krisp →

4. LALAL.AI

• Platforms: Web, Windows, macOS, Linux Ubuntu.

This simple service uses a neural network to remove background noise from audio files uploaded for processing. LALAL.AI Voice Cleaner designed for human speech and highlights exactly the voices in bad recordings. And Stem Splitter can split music into separate groups of instruments and cut vocals from songs.

When adding a file for noise removal, you can only set the filtering level - soft, normal or aggressive. In free mode, LALAL.AI will only process 10 minutes of audio up to 50MB. For those who improve recordings regularly, there are paid plans: $15 for 90 minutes and $25 for 350 minutes.

Try LALAL.AI →

5. Adobe Enhance Speech

• Platforms: web.

There are no settings in Adobe Enhance Speech. The neural network neutralizes the echo and removes most extraneous noise automatically. You can upload a file up to an hour long and up to 1 GB in size, and then compare the original entry with the filtered one.

If the voice is recorded too quietly compared to the surrounding sounds, then some of the words may disappear after processing. Enhance Speech not only reduces the amount of noise, but also restructures the speech itself to improve its clarity.

The algorithm copes well with the English language, but in other cases a funny situation occurs. For example, Russian may sound like a person is speaking with an American accent. These oddities do not appear in all words, but it is worth remembering this feature of the service.

Enhance Speech as part of the Adobe Podcast suite is currently available for free.

Try Adobe Enhance Speech →

6. Audio Noise Reducer

• Platforms: iOS, Android, web.

The mobile app allows you to record voice notes or even podcasts on the go. In addition, you can upload a ready-made file weighing up to 50 MB to the program for processing. Audio Noise Reducer does a good job of filtering background noise, and it works not only with audio, but also with video.

Audio Noise Reducer is available for free, but with ads. It can be removed by subscription for 269 rubles per week, 890 rubles per month, or a one-time fee of 3,090 rubles. This cost also includes additional formats for saving.

Try Noise Reducer Web →