Trusting what is and isn’t real on the internet has become harder than ever. With AI everywhere, the lines have become blurrier when determining what is and isn’t a real voice. Thankfully, a few certain signs can give away everything you need to know.
6No Natural Pauses
In some cases, no natural pauses are a good way todetermine if a video is AI-generated—and the same goes for audio. Newer and more expensive models tend not to have this problem, but earlier versions of AI and cheaper models aren’t on the same level.
For example, you may’t hear any natural pauses in the AI-generated podcasts below. While both voices definitely sound human-like, including elements like breathing, neither person stops when it makes sense to do so. Moreover, some parts of the podcast involve one of the hosts not finishing their sentences before the other speaks.
Of course, no natural pauses can also be the result of bad editing. But in many cases, it’s the result of earlier versions of AI software. Even though people may still create with these models today, you should especially use this tip for content published in 2023 or earlier.
5Random Glitches in the Voice
AI audio creation has come a long way, and nowadays, it’s easier than ever to generate something that resembles real life more accurately. But while this is the case, AI is still not perfect. As a result, you can often notice glitches and other errors if you listen closely enough.
If you’re listening to multilingual audio, glitches become even more evident. For example, the “gracias” in the video below—which was generated using AI—isn’t pronounced correctly. In similar videos, you’re able to sometimes notice glitches if the accent is from someone speaking in their non-native language.
Like the previous tip, this isn’t as relevant for more advanced LLMs. But for those without significant development, you’re more likely to notice such issues.
4Inconsistent When Put on Speed Play
Playing content at a faster speed than it was recorded isn’t just an optimal way to absorb information more quickly in some cases. If you’re trying to determine whether the audio you’re listening to was generated via AI, putting the content at 1.25x speed or higher can make it pretty evident that you’re not listening to a real human voice.
When you play AI-generated audio at a faster speed than it normally is, the content often sounds quite robotic. It’ll feel like you’re listening to a meme at times rather than something genuine. Similarly, you may notice inconsistencies in sounds when you listen to AI-generated music.
Even though you cancreate an AI voice that sounds like youthese days, it’s still not going to be on the same level as if you spoke yourself.
3No Breathing
In addition to no natural breaks, another potential sign you’re listening to AI audio—especially if it’s in a podcast or video—is that you can’t hear any breathing. Even if something has been heavily edited, you will often hear the person breathing if it’s their voice and not generated by an AI tool.
Besides no breathing, you won’t hear people taking breaths or anything similar while they’re actually speaking. For example, I have rarely heard someone sigh when the voice is AI-generated The video below, which was developed using AI, is one example.
While AI is getting better at replicating human interactions, it still sometimes feels like it has no soul. You’ll probably notice this as you listen to more AI-generated content over an extended period. This is one of many reasons whyAI-generated podcasts are polarizing(and probably will stay that way for the foreseeable future).
2The Voice Is Overemotional
While AI voices are more capable of sounding like people than before, they aren’t necessarily great at portraying the full range of human emotions. One criticism has been that AI voices sometimes sound robotic, but the opposite is also true. The short movie below, created with AI, is one example.
In many instances, AI voices are overly emotional. For example, they’ll exaggerate how excited they are to the point that it often sounds forced. Even a basic story will sometimes sound like the most amazing thing ever, to the point that even someone with genuine passion wouldn’t speak like that.
If you suspect that the voice doesn’t sound genuine, it’s often because either someone is not authentic or the voice was generated with AI.
1Your Intuition Says It’s AI
Our gut feelings have evolved to keep our species alive, and while we’re often told to ignore them in the modern day, they’re usually right. While you can run comprehensive analyses and listen to something multiple times to try and decode whether it’s AI-generated, you sometimes just have a feeling.
If you’re on the fence about whether what you’re listening to was generated by AI, you’re probably right. You’ll become more attuned to this the more you listen to AI-generated content, such as if youspend a week listening to AI-generated music.
While difficult, you can use several tactics to decipher whether you’re listening to an AI voice. Comparing how it acts compared to ordinary human beings is one way, and you can also look for glitches. Sometimes, though, your intuition is the best judge.
