Unless you’ve been living under a rock, you’ll have heard of ChatGPT. It’s fair to say that we’ve seen a seismic shift in what AI is able to do, to the point where the heads of national governments are genuinely holding meetings discussing whether this new AI tech is a risk to humanity.
Only time will tell whether AI does ultimately become self-aware and destroy its human overlords, but in the meantime, it’s a whole lot of fun to play with. You’ve been able to use the ChatGPT app on your phone for quite a while, and it’s completely free if you only want to use the standard version. But now there’s a whole new way to interact with ChatGPT: ChatGPT Voice.
As you’ve probably already guessed, ChatGPT Voice allows you to actually talk to your AI chatbot, rather than having to type everything. And ChatGPT can now talk back, using impressively realistic AI voices. Here’s what you need to know about ChatGPT Voice.
Is ChatGPT Voice free?
When ChatGPT Voice first launched back in September 2023, it was only available to people with ChatGPT Plus subscriptions, or Enterprise users. However, on 21 November 2023, Greg Brockman, one of the co-founders of OpenAI, posted on X that the feature was being rolled out for all free ChatGPT users, too. Provided that the feature has been rolled out to your device, you’ll be able to use ChatGPT Voice without needing to pay for a subscription.
How to use ChatGPT Voice
Using ChatGPT Voice is incredibly easy to do. All you need is the ChatGPT app installed on your phone, and you’re good to go.
- Launch the ChatGPT app.
- In the bottom right of the screen, tap the headphones icon.
- The first time you open ChatGPT voice, you’ll be prompted to choose from one of the five voice options.
- Select your voice of choice and tap Confirm. You can change your voice selection at any time in the settings.
- A white circle will appear and bounce from side to side on your screen.
- Once it’s finished bouncing, ChatGPT Voice is ready to go.
- Start speaking. ChatGPT will respond when it thinks you’ve finished talking.
- If you find that ChatGPT interrupts you before you’ve finished speaking, tap and hold the screen whilst you’re talking. ChatGPT will then only respond once you let go of the screen.
- Once ChatGPT has responded, you can continue your conversation; the AI will remember everything that came earlier in the conversation, allowing you to have a natural back-and-forth conversation.
What can ChatGPT Voice do?
The short answer is that ChatGPT Voice can do pretty much everything that ChatGPT can do through the text interface, without all that tedious typing. I asked ChatGPT Voice what it was capable of, and the response was that it could ‘answer questions, provide info on various topics, perform language translation, and even do a bit of storytelling.’ And that’s a fairly accurate summary.
You can ask ChatGPT voice for information, although there are limitations, which we’ll look at in the next section. Unlike other voice assistants, such as Siri or Alexa, however, you can drill down into the information by asking follow-up questions. For example, if ChatGPT tells you that the biggest prehistoric shark was the Megalodon and that it could reach lengths of up to 20 meters, you can follow up by asking it to put that size into context based on real-life objects. ChatGPT Voice gave me a response that it’s equivalent to two school buses parked end to end, which is a fairly accurate estimate. Try doing that with Siri, Alexa, or Google Assistant and you’ll get nowhere fast.
Language translation is simple to do; just ask ChatGPT to translate a phrase into another language, and you’ll get a voice response of the translated phrase. This works pretty well with common languages, such as Spanish, German, and French, with the voice speaking the foreign phrase with a fair attempt at the correct accent. But when I tried the same with Swedish, the voice read the translation phonetically, which sounded nothing like the actual Swedish pronunciation, so there are clear limitations. You probably shouldn’t end your Duolingo streak just yet.
You can ask ChatGPT to tell you a story and choose from a choice of story styles. You’ll then be read a supposedly original story, which is actually quite soothing. ChatGPT told me a short story about a village where the clock in town would ring an extra chime each night. I was able to ask the AI to expand on the story, and it added more details to the story. It wasn’t the greatest story I’ve ever heard, but it’s still quite an impressive feat of technology, and potentially a great way to get your kids to sleep, as long as the story doesn’t take too dark a turn.
What can’t ChatGPT Voice do?
Whilst ChatGPT can do a lot, it has some significant limitations. Probably the biggest limitation is that the free version of ChatGPT was trained on huge amounts of data, but that data only went up to January 2022. If you want to know about anything that happened beyond that date, you’re bang out of luck. Aliens might have landed and taken over the Earth in February 2022, and ChatGPT wouldn’t have a clue about it.
In real terms, this means that there are a lot of things that Siri, Alexa, or Google Assistant can do that ChatGPT Voice can’t. For example, you can ask Siri for the latest football scores, and she can provide that information in an instant. Ask ChatGPT voice, and it will tell you that it doesn’t have real-time information and that you should look online. Not ideal for a virtual assistant.
ChatGPT Voice also won’t engage in harmful or illegal activities, provide personal information, or produce inappropriate content. And possibly most importantly, it won’t always give you accurate information. I tried asking quite a complex question about a crossword clue that had me stumped. The clue had a two-word answer made up of a five-letter word and a four-letter word. The first letter of the second word also had to be the same as the third letter of the first word. Despite trying numerous times to request that the answers had to meet all these criteria, the responses would constantly fail to meet one or more of the stipulations, giving me words that were too long, or words that didn’t start with the appropriate letters. When pressed, it eventually resorted to giving answers that were total nonsense, although often quite amusing.
A final thing that ChatGPT Voice really can’t do is be funny in an original way. If you want a real laugh, try asking ChatGPT Voice to come up with some ‘original’ Christmas cracker jokes. The results are quite encouraging if you were previously worried about AI taking over the world, because the current models can’t even write cracker jokes to the level of an eight-year-old. Your job may still be safe. For now…
What do the ChatGPT Voice voices sound like?
Honestly, they’re excellent. There are five voices to choose from, and all of them sound impressively natural. OpenAI says that they’re built with a new text-to-speech model that can generate human-like audio from just a few seconds of sample speech. Each voice is based on speech recorded by a voice actor, and you wouldn’t know that the phrases were generated rather than pre-recorded.
You can even ask ChatGPT to respond to you in a sarcastic tone, and it will do exactly that, giving a really impressive sarcastic tone of voice. Even Elon Musk would be impressed, seeing as this is the supposed USP of Grok.
The ability to create natural sounding voices from just a few seconds of sample speech obviously has some rather sinister connotations; but it’s also quite cool that the tech is there to mimic the voice of someone else, just like in Mission Impossible. Now all we need are prosthetic faces.
How to replace your voice assistant with ChatGPT Voice
There’s a lot that ChatGPT can do that’s much better than traditional voice assistants, including the ability to have a back-and-forth conversation without having to repeat the key information. If you want to replace the voice assistant on your phone with ChatGPT Voice, there are a few ways to do so.
How to launch ChatGPT Voice with the Action button on iPhone
This is a great way to use ChatGPT Voice, as you can still use Siri for queries that require access to current information, but use ChatGPT when you want to get deeper information on topics that don’t rely on current data.
- Open the Settings app.
- Scroll down and tap Action Button.
- Swipe left to the Shortcut option and tap the currently selected shortcut at the bottom of the screen.
- Scroll down to the list of apps, and you should see a ChatGPT: Start Voice Assistant option.
- Select this option.
- Now, when you press the Action button, ChatGPT Voice will launch.
How to launch ChatGPT Voice with Siri
If you don’t have an iPhone 15 Pro or Pro Max, you won’t have an Action button to invoke ChatGPT Voice. In this case, you can launch the app using Siri by using the Shortcuts app. You can do a similar thing using the Tasker app on Android, too.
- Open the Shortcuts app and tap the plus (+) icon in the top-right corner of the screen.
- Tap Add Action.
- In the search bar, type ChatGPT.
- In the search results, tap the Start voice conversation option.
- Tap the arrow next to the name of your shortcut and select Rename.
- Give the shortcut a name that you can speak, such as Your Time is Over.
- Tap Done to save your shortcut.
- Now all you need to do is say ‘Hey Siri, your time is over’ and ChatGPT Voice will launch.
- You can still ask Siri for information in the usual way.
What devices does ChatGPT Voice work on?
Currently, ChatGPT Voice is available on the Android and iOS ChatGPT apps. It remains to be seen whether the feature will roll out to other devices. It would be great to be able to use ChatGPT Voice on an Apple Watch, for example, as the Siri experience on Apple Watch is far from ideal, although things have improved with the new Apple Watch 9 and Apple Watch Ultra 2 that use on-device processing to add some useful new options.