Better Scripts, Better Audio: Writing for Human and AI Voices

If you produce a lot of video content, you might find yourself writing scripts for the audio portion. Here are a few quick tips to help you write natural-sounding and engaging scripts. 

You want your audio to sound natural, so you should write like you speak. Here are some suggestions: 

While this is a good rule to follow in any writing, repetition in an audio script can be more noticeable than it is in writing.  You don’t want to use the same words over and over or start each sentence in the same way, or your audio will sound monotonous. Try mixing up constructions and sentence structure to keep things interesting, just as you would with good writing. 

Sometimes it’s hard to pick up on elements like contractions or repetition when you are writing, so I find it helps to read your text out loud as you review it (this is a good hint even if you aren’t writing a script!). This will help you process the content differently and allow you to pick up on some of the nuances in your script that you might otherwise skip over when reading. Also, if it’s hard to say, it will be hard to record, so consider revising if you find yourself stumbling over any phrases. 

As an alternative (or addition!) to reading your text out loud, some programs have tools to let you listen to your script, like the Read Aloud tool in Microsoft Word (Review tab>Speech panel). Listening to your script is a great way to ensure it flows naturally. 

Tips for AI-Generated Audio 

Once you have a script, recording the audio can be a time-consuming process, as it often will involve editing out flubs or adjusting for background noise. Programs like Audiate allow you to input a script and generate an audio file using one of their AI voices or avatars. While this can save time, there are some nuances to be aware of when it comes to creating your scripts. 

While I’ve found that the generated audio is fairly accurate, it doesn’t always get the pronunciation quite right. This is especially true for certain brand names, acronyms, or words that can be pronounced multiple ways. Listen carefully to the audio and make adjustments as needed. Here are some tricks I’ve used: 

You should also listen carefully to ensure your sentences have the proper inflection. For example, the pitch of the voice should go up at the end of a question. You can use similar tricks with hyphens or spaces to try to change how the program “reads” the text, but this one is a bit harder to adjust so you may have to play around with your particular program’s settings. 

Once you’ve selected an avatar for your voice, you will need to adjust the audio speed accordingly. Pick something that sounds natural to you, without being too fast or too slow, then listen for any instances where the pacing seems off – you can often adjust certain parts by adding extra spaces if the words are being said too quickly. 

Also take the purpose and audience of your video into account – you may want a slightly slower pace for your audio for a training video vs. a more casual marketing piece, for example. 

I hope these tips help you with your scripts, regardless of how you generate the audio!