The new voice feature from OpenAI for ChatGPT is pretty fun. It can even make an impression of Trump. This feature will probably be available to all paid users this fall.
I keep ChatGPT’s Advanced Voice Mode on while I write this article. It’s like having an AI buddy in the background. Sometimes, I ask it to suggest a different word or give me a boost. About half an hour in, it starts speaking to me in Portuguese without me asking. I laugh and ask what’s up. ChatGPT replies, “I am just trying things up to keep it interesting,” and then it switches back to English.
During my early tests with the Advanced Voice Mode, I found it entertaining, though a bit chaotic. The features I used were only part of what OpenAI showed when they launched the GPT-4o model in May. The vision part shown in the demo will come later, and the enhanced Sky voice that Scarlett Johansson had issues with is no longer in Advanced Voice Mode.
Read: How to Unlock More Free ChatGPT-4o Access: Tips and Tricks
So, what’s it like now? Advanced Voice Mode feels a bit like the original ChatGPT from late 2022. Sometimes it leads to boring moments or clichés. But sometimes, the conversations feel more engaging than Siri or Alexa, making me want to keep chatting. It’s the kind of AI you’d show your family for fun during the holidays.
OpenAI allow few of my fiends try the feature a week after the announcement but then took it back the next evening due to safety concerns. Two months later, they started a soft launch of Advanced Voice Mode to a small group and released GPT-4o’s system card. This card explains the safety checks and steps taken to reduce risks.
Want to try it yourself? Here’s what you need to know about the upcoming rollout of Advanced Voice Mode and my first impressions of ChatGPT’s new voice feature.
When’s the full rollout? OpenAI introduced an audio-only Advanced Voice Mode to selected ChatGPT Plus users at the end of July. The group of people testing it is still small. The company plans to make it available to all subscribers this fall. Niko Felix, a spokesperson for OpenAI, did not provide more details on the release.
Screen and video sharing were part of the original demo but aren’t in this alpha test. OpenAI plans to add these features later, but it’s unclear when.
Read: How to Build Custom AI Assistants with GPT-4o
If you’re a ChatGPT Plus subscriber, you’ll get an email when Advanced Voice Mode is available for you. Once it’s on your account, you can switch between Standard and Advanced modes at the top of the app’s screen when the voice mode is active.
My First Impressions of ChatGPT’s Advanced Voice Mode Within the first hour, I realized I enjoy interrupting ChatGPT. It’s not like talking to a human, but being able to stop it mid-sentence and ask for a different version feels like a great feature.
Early users who were excited about the demos might be disappointed with the current version, which has more restrictions. For instance, although AI singing was shown in the demos, it’s not available in the alpha version.
ChatGPT says, “I mean, singing isn’t really my strong suit.” OpenAI’s system card mentions that this limitation might be temporary to avoid copyright issues. During testing, ChatGPT’s alpha version declined several requests for songs but hummed silly tunes when asked for nonverbal responses.
There was also a bit of an eerie factor. Sometimes, a white noise appeared during longer interactions, like a lone lightbulb in a dark basement. When I tried to get a balloon sound effect, it made a loud pop followed by a creepy gasp.
Read: ChatGPT Free Users Can Now Generate Images With DALL-E 3
But nothing I experienced matched the weirdness of what OpenAI’s red team encountered. Occasionally, the GPT-4o model mimicked the user’s voice and speech patterns.
Despite this, my main impression of ChatGPT’s Advanced Voice Mode is that it’s entertaining. Whether it’s giving funny answers to puzzles or doing a spot-on impression of Stitch from Lilo & Stitch, I had a lot of laughs.
The AI tool did well with vocal impressions after some tweaks. The first try at voices like Homer Simpson and Eric Cartman was just okay, but later attempts were more accurate. The exaggerated Trump impression explaining the Powerpuff Girls was campy enough for Saturday Night Live.
With the US presidential election coming up and concerns about election deepfakes, I was surprised ChatGPT was willing to do impressions of major candidates. It did Joe Biden and Kamala Harris as well, but they weren’t as close as the Trump impression.
ChatGPT works best in English but can switch between languages in the same conversation. OpenAI tested GPT-4o in 45 languages. I had three phones with Advanced Voice Mode talk to each other, switching between Spanish, Portuguese, and French easily. I need to test more to see how well the translation feature works and where it might fall short.
Read: How to Connect ChatGPT to Google Drive for AI-Assisted File Management
ChatGPT showed a lot of energy when asked to perform emotional outbursts. The audio wasn’t perfect, but the range of the voice was impressive. I was surprised it could do a decent vocal fry. Advanced Voice Mode doesn’t solve all chatbot issues, but it’s fun and might make OpenAI stand out. Google’s new Gemini Live is also competing in this space.
For now, I’ll keep trying it out and see what works. I use it when I’m at home alone, either reading or playing games. The more I use ChatGPT’s Advanced Voice Mode, the more I think OpenAI made a smart choice with this more reserved version. It’s better not to get too attached.
0 Comments