{"id":286598,"date":"2025-11-16T10:55:38","date_gmt":"2025-11-16T10:55:38","guid":{"rendered":"https:\/\/www.newsbeep.com\/ca\/286598\/"},"modified":"2025-11-16T10:55:38","modified_gmt":"2025-11-16T10:55:38","slug":"the-ai-accent-is-an-easy-way-to-spot-ai-videos","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/ca\/286598\/","title":{"rendered":"The ‘AI Accent’ Is An Easy Way To Spot AI Videos"},"content":{"rendered":"

Do you know what artificial intelligence<\/a> sounds like? When asked to guess, most people can\u2019t tell the difference between AI-generated voices and real human conversation, according to multiple<\/a> studies<\/a>.<\/p>\n

This confusion can have disastrous consequences on how we see the world. When you get confused about what\u2019s real or not on screen, you can start to believe misinformation, and in worst cases, racist stereotypes about people being depicted in AI-generated videos<\/a>. <\/p>\n

But there might be one reliable way to suss out what\u2019s AI, especially on video: Listen to how the people sound. <\/p>\n

A range of AI experts shared the telltale signs of why the voices and sounds in an AI video can often reveal its synthetic origin. Here\u2019s how. <\/p>\n

\"AI<\/p>\n

Illustration: HuffPost; Photos: Getty<\/p>\n

AI voices in Sora videos often sound like they have downed five cups of coffee. <\/p>\n

Listen for the over-caffeinated tone. <\/p>\n

Real people have a natural rhythm to how they speak, so that some words are said more slowly than others. But AI voices often sound unnaturally rushed all the time. <\/p>\n

Jeremy Carrasco, a video expert who debunks AI videos on social media, said he notices that videos from Sora \u2015 an artificial intelligence video app owned by OpenAI \u2015 often have an \u201coverly energetic\u201d quality. \u201cThey\u2019re saying so much and they\u2019re not saying much at all, they\u2019re just cramming in words,\u201d he said. <\/p>\n

Even OpenAI is aware of this telltale sign. Too many em dashes in a text answer is known<\/a> to be a giveaway in OpenAI\u2019s ChatGPT answers that can reveal when someone\u2019s cover letter or first date message got AI-generated.<\/p>\n

In October, the hosts of video streaming show TBPN asked Bill Peeples, the head of Sora, about what the \u201cem dash of [AI] video\u201d was in an interview<\/a>. His immediate response was telling.<\/p>\n

\u201cI think right now the \u2018em dash\u2019 is this slightly wired speech pattern in Sora where it likes to say a lot of words quickly,\u201d Peeples said<\/a>. <\/p>\n

Watch out for garbled, slurred voices. <\/p>\n

What we might call someone\u2019s speaking rhythm is what linguists would call \u201ccoarticulation,\u201d or how our voices physically go from one sound to another as air goes through our noses and out our mouths. And a lot of AI-generated speech is still bad at this and makes garbled sounds that appear to flatten out natural sound pitches. <\/p>\n

\u201cNo human being would ever produce that same kind of garbled quality [as an AI-generated voice], because, literally, we can\u2019t,\u201d said Melissa Baese-Berk, a linguistics professor at the University of Chicago. \u201cOur vocal track can\u2019t go from one sound to another without some blurring of the information between those two sounds.\u201d<\/p>\n

Baese-Berk used the example of an AI subway meet-cute video<\/a> where a woman meets a man she immediately calls her \u201chusband.\u201d The video fooled many people into believing it was real. But when the woman says \u201chusband,\u201d the \u201cband\u201d part of the word sounds \u201csuper duper weird,\u201d she said. The \u201cband\u201d part of the word \u201cis missing the natural coarticulatory information that happens when you move from the tip of your tongue to your lips,\u201d Baese-Berk said.<\/p>\n

\u201cOnly a robot could go from their tongue to their lips without having any kind of mashing up of those sounds,\u201d Baese-Berk said. <\/p>\n

This inhuman mash-up of words is by design. <\/p>\n

\u201cText-to-speech models are trained to predict the most likely pronunciation of a word in sequence, but they often struggle to smoothly blend the sounds that connect words,\u201d said Mig\u00fcel Jett\u00e9, vice president of AI at Rev, a speech-to-text service. \u201cFor example, where a human might naturally say \u2018didja\u2019 instead of \u2018did you,\u2019 AI has a tendency to either over-enunciate each word, or blend them too abruptly.\u201d<\/p>\n

Pay attention to mispronounced words. <\/p>\n

If there\u2019s an obviously mispronounced word, that can also be a sign, Jett\u00e9 said, because \u201cAI voices can struggle with unusual or unique words that don\u2019t appear in the training data.\u201d <\/p>\n

Google<\/a>\u2019s text-to-video Veo model, for example, \u201cmight not be cramming in as many words, but they will put them out of order, or the wrong person will say something,\u201d Carrasco said he has observed. <\/p>\n

Notice when emotional reactions don\u2019t match the story of the video. <\/p>\n

In a 2025 study<\/a> that asked participants to rate which voices were AI or not, the AI voices created by text-to-speech models were only identified accurately 55% of the time. The biggest mistakes happened with AI voices that sounded angry.<\/p>\n

This may be because participants expected AI voices to sound robotic, said Camila Bruder, a co-author of that study and a researcher from the Max Planck Institute for Empirical Aesthetics.<\/p>\n

In reality, AI voices are often too emotional for what the scene calls for. If the AI voice is \u201ctoo stereotypically happy, like, \u2018Wow!\u2019 or it\u2019s stereotypically mad…like a bad actor,\u201d these traits can be indicators that the video is AI, Bruder said. <\/p>\n

Carrasco said you should also notice when what is being said is an odd emotional reaction. Take one viral AI video<\/a> of fish falling from the sky. \u201cThey\u2019re fish, they\u2019re actually fish!\u201d a woman in the video exclaims.<\/p>\n

\u201cThey\u2019re just narrating what\u2019s happening on the screen. You wouldn\u2019t do that in real life,\u201d Carrasco said about this video. \u201cIf a bunch of fish were raining [down], I\u2019d probably just say \u2018What the fuck.\u2019\u201d <\/p>\n

Compare the inappropriate AI emotions to the real-life horror a truck driver recently experienced when he was filmed<\/a> watching a plane crash that happened in front of him in Kentucky. In this video, the driver doesn\u2019t narrate his experience, his mouth simply drops open. \u201cHe\u2019s just in disbelief. That\u2019s kind of how a lot of these would be\u201d in real life, Carrasco said.<\/p>\n

You can also simply look at what people\u2019s mouths are doing for clues. \u201cThe visual giveaways in these videos can be just as revealing as the audio,\u201d Jett\u00e9 said. \u201cIf the speaker\u2019s lips don\u2019t perfectly sync with the audio…that\u2019s a strong indicator.\u201d<\/p>\n

These clues are helpful, but they aren\u2019t always guaranteed. <\/p>\n

PowerTheTruth<\/p>\n

Your SupportFuelsOur Mission<\/p>\n

Your SupportFuelsOur Mission<\/p>\n

Join HuffPost members who keep fearless, inclusive journalism alive.<\/p>\n

We remain committed to providing you with the unflinching, fact-based journalism everyone deserves.<\/p>\n

Thank you again for your support along the way. We\u2019re truly grateful for readers like you! Your initial support helped get us here and bolstered our newsroom, which kept us strong during uncertain times. Now as we continue, we need your help more than ever. We hope you will join us once again<\/a>.<\/p>\n

We remain committed to providing you with the unflinching, fact-based journalism everyone deserves.<\/p>\n

Thank you again for your support along the way. We\u2019re truly grateful for readers like you! Your initial support helped get us here and bolstered our newsroom, which kept us strong during uncertain times. Now as we continue, we need your help more than ever. We hope you will join us once again<\/a>.<\/p>\n

Support HuffPost<\/a><\/p>\n

Already a member? Log in to hide these messages.<\/a><\/p>\n

Of course, these clues are not always a guaranteed way to reveal an AI-generated voice. ElevenLabs, the AI lab which clones real voices, is good at adding vocal fry and human pauses, so listening for a voice that speaks without breaths isn\u2019t \u201calways the case\u201d that it\u2019s AI, Bruder said. <\/p>\n

But as a whole, these telltale signs are a strong indicator that the video you are watching was probably created by a machine. And that\u2019s a helpful start. As AI continues to evolve at breathtaking speeds, we need all the help we can get to understand what\u2019s fake and what\u2019s not. <\/p>\n

\u201cIf something feels off, it probably is,\u201d Jett\u00e9 said. \u201cA healthy dose of skepticism and a good eye and ear for detail can go a long way.\u201d<\/p>\n