Maybe you experienced it in the past: you've taken an interview and you want the audio to be converted to text automatically - but the result is terrible. Or worse: the interview is not even audible.
The extent to which your audio can be automatically converted into text is directly related to the quality of your audio. Perfect quality audio can be converted into text up to 98% automatically. Audio of very poor quality sometimes cannot be converted automatically at all.
Below you can read in 4 steps how to ensure that your audio files are of high quality, to improve your automatic transcripts:
How do you record good quality audio?
Keep a small distance between the speaker and the microphone
Make sure there is as little background noise as possible
Do not interrupt each other
Provide a good quality microphone
Below we will discuss the above points in more detail.
1. The distance between the speaker and the microphone
The first and the easiest tip is the distance between the speaker and the microphone. Sound quality is greatly influenced by the distance from the speaker to the microphone. If the distance is too far, the microphone will not pick up the sound properly and parts may get lost. Too far a distance between microphone and speaker is the main cause of poor automatic transcriptions. If the distance is too close, you will hear the speaker breathing into the microphone all the time (not a pleasant sound). The perfect distance depends on the microphone, but often the ideal distance is around 10 centimetres. Recording with a telephone at 50 centimetres from the speaker is therefore not advisable for sound quality. There are microphones that enable a much farther the distance from the speaker, but with most normal microphones it is wise not to keep them too far away from the speaker. Holding the telephone in front of the speaker's mouth gives the best result. If that is not desirable, the phone should be placed on the table right in front of the speaker. What is even better and more pleasant is the use of a separate microphone, see point 4.
Another tip for people who record sound with their phone: take your case off. Removing the cover makes sure that the microphone is visible and can record more and better sound.
2. Eliminate background noise
The audio quality improves enormously when most of the background noise is eliminated.
Technically it is never completely silent, even in the quietest environments there will be some sounds. It is therefore impossible to have complete silence anywhere, but we do have a few tips to make sure it is as quiet as possible at the moment of recording:
Make sure everyone around you is quiet. Tell everyone around you that you are conducting an interview and that it is important there is silence. Put a sign on the door so nobody just enters without asking.
Turn off all devices that may produce any sound. This could be the air conditioning, a fan, a printer or your computer. It is not enough to just turn off the sound on your computer as most computers have built-in fans that produce sounds. If you need your computer during the interview, it is important to keep it as far from the microphone as possible.
3. Do not interrupt each other
Besides that, it is not quite correct to interrupt each other, this also confuses our transcription software. That is why it is important to let each other finish your sentences so our software can do its job as well as possible. Furthermore, our software has some trouble with heavy accents. Of course, it is difficult to do something about an accent, but it is an enormous help to try to talk as accent-free as possible. By following these tips, we can best convert your audio into text.
4. Quality of the microphone
Of course, the quality of the audio depends greatly on the quality of the microphone. With microphones it is often true that more expensive microphones also have better quality. Fortunately, there are also cheap microphones for recording good quality audio, if only because they ensure that the speaker and the microphone are at the right distance from each other.
There are different types of microphones, these are:
- A lavalier (clip-on microphone)
The lavalier is excellent for interviews. Lavaliers provides better recording quality than a telephone and ensures that there is an optimal distance between the speaker and the microphone. Another advantage is that lavaliers are often equipped with two heads, so you do not need to move the microphone for another speaker during an interview.
2. A table microphone
A table microphone is a microphone that can be placed on the table between the speakers. Most table microphones provide better recording quality than a telephone. A table microphone can be useful in a conversation with several speakers. If the speakers are in different directions of the microphone, it is important that the polar pattern of the microphone is Omni Directional (more on this later).
3. A telephone
Currently there is voice recording software available for almost every smartphone. Because of this the smartphone is the most accessible microphone in this list. The recording quality is pretty good for most smartphones, but it does improve the quality if you keep the phone close to the speaker and do not put it on the table.
4. A laptop
Most laptops have a built-in microphone, this microphone can also be used for recording interviews. The laptop and phone are both good options if you do not want to invest in a separate microphone. However, for recording interviews, we recommend using a smartphone over a laptop. This is because a smartphone is easier to use and records in better quality audio.
5. A voice recorder
A voice recorder, often called a dictaphone, is an excellent microphone to be used for recording interviews. The quality of the recordings is usually very high in a voice recorder. Also, with a voice recorder, the quality of the recording improves enormously when you keep the voice recorder near the speaker instead of putting it on the table.
There are different types of microphones that record sound from different directions (also called polar patterns).
For interviews, a Figure-of-Eight is the best choice, because it absorbs the same amount of sound on both sides of the microphone. A Figure-of-Eight also records less background noise than an Omni Directional. However, Figure-of-Eight microphones are often a lot more expensive.
Research on the accuracy of our software with different microphones reveals major differences, as can be seen in the table below. Word Error Rate means the percentage of errors that our software makes. For example, a Word Error Rate of 10% means that there are on average 10 errors for each 100 words. Find the results of the research below:
|Microphone||Word Error Rate|
|Lavalier (a clip-on microphone) + telephone||10% - 20%|
|Premium telephone on the table||20% - 30%|
|Premium telephone in your hand||15% - 25%|
|Budget telephone on the table||30% - 80%|
|Budget telephone in your hand||20% - 50%|
|Laptop||30% - 50%|
|Table microphone||10% - 40%|
As you can see above, there is already a significant difference between putting your phone on the table and holding it in your hand. Furthermore, the quality of your microphone is extremely important. With a lavalier, the quality of the audio can be significantly improved, and the Word Error Rate reduced.
We have selected lavalier and table microphone options with high sound quality for you, which you can use with your smartphone:
Budget lavalier: Maono AU404
Budget table microphone: Audio Technica ATR 4697 and a TS to TRRS connector
Everything at a glance
As discussed above, it is actually very simple to improve the quality of your audio files. First, it is important to keep about 10 centimetres between the speaker and the microphone. Secondly, it is important to eliminate background noises, so unplug any devices that can produce sound and tell everyone near you that they should be quiet. Thirdly, it is important not to interrupt each other. Finally, the quality of the sound depends greatly on the quality of the microphone and the type of microphone.
If you follow these tips, the quality of your audio files will improve enormously, so we can better convert them to text and you have to make less and less adjustments to your automatically transcribed texts!