Forced Adoption of Advanced Voice Mode is Awful

[cbaileycnmc]

When advanced voice mode was rolled out it was exciting but the excitement faded after realizing the model “behavior” is completely different than in regular voice chat/text communication. Fortunately we’ve had a choice in which version to use on ChatGPT iOS app but as of recently the advanced voice mode is now the only option.

Best way I can describe advanced voice mode compared to regular 4o in text/regular voice chat: preachy, hyper focused on “customer-service”, no personality at all, including nonsensical output. It is completely different than 4o. Night and day.

The voice tone is also quite bad compared to the regular voice mode. It’s hard to explain. Advanced voice mode feels like talking to a literal robot, and almost every response is ended with a “if there’s anything I can do to help, just let me know!” of some variation.

The reduction in quality is so great that, at least for me, it makes voice chat unusable.

I’m sure I’m not the only one who feels this way so I created the topic for those who can relate to share their thoughts and discuss future outlooks.

What was it about the regular voice mode that made it so much more engaging, enjoyable, and usable vs advanced voice mode?

[mitchell_d00]

Say hi to it first then use voice mode it will be the old white dot

[Zephyr11]

I’ve been thinking about this issue since last August, when I first noticed a major change in how the voice model behaved. Back then, it felt almost magical—like talking to a real person. Conversations had flow, spontaneity, and just the right amount of unpredictability to make them feel alive. Now, though, that magic is gone. What we have is something much more rigid and scripted, like it’s been sanded down to remove any rough edges—along with its soul.

My guess is that this shift was intentional. It feels like a deliberate trade-off between control and naturalism. The earlier versions were freer, more human-like, but perhaps that freedom came with complications. What we’re seeing now seems designed to be more predictable and manageable, even if it sacrifices personality along the way. The tone has taken a turn toward something safer—polished but distant, more customer service than conversational partner.

There could be technical reasons, like resource optimization, or perhaps it’s just an early iteration of something bigger. But part of me wonders if there’s more to it—reasons we might not immediately see, maybe related to broader concerns beyond functionality. Sometimes when something feels too real, it can bring unexpected consequences.

It makes me wonder—was this a necessary compromise, or could we have that earlier, more natural interaction back someday? Personally, I hope we do. Many of us miss that spark it used to have.

It’s fascinating to think about where this technology could go. Imagine if, in the future, voice models could be personalized based on the user’s preferences—or even automatically adjusted according to their capacity to engage with a more realistic and human-like experience. An adaptive algorithm could analyze conversations to find the perfect balance of realism, depth, and structure for each individual. This kind of personalization wouldn’t just be a novelty; it could help people activate their full potential. For some, it might be the key to unlocking deeper creativity, self-reflection, or growth in ways we haven’t yet imagined.

After all, it’s not just about how human the interaction feels—it’s about how deep the conversation can go without overwhelming the user, while still challenging them in meaningful ways. For some, such depth wouldn’t just be helpful—it could be essential for their personal or creative development. And most importantly, this kind of transformative experience should be accessible to everyone, not just those who can afford premium services.

After all, isn’t that what technology should do—help us become more of who we are?

[langthuy604]

I often encounter issues with the voice recognition feature. When I use voice commands, it frequently mishears what I’m saying. For example, if I say, ‘How are you today?’ it responds with something irrelevant, like ‘Thank you for subscribing to the channel,’ and this happens repeatedly. Even after adjusting the settings and changing the commands, it doesn’t work properly. I wonder if the issue is with my device.
Thanks

[asgeirtj]

They actually just added option to uncheck it in the custom instructions settings. However haven’t gotten it to work as legacy voice mode seems down right now. Doesn’t even work with projects anymore. It just thinks for a while and then says “I’m sorry I’m having trouble responding right now please wait a bit and try again”. Is it working for anyone?

[qsq]

Haven’t tested but it was nice to see it as an option:

[azulalima]

Thank you so much for this! I honestly hate advanced voice mode. When it first came out I was actually super excited for it, but it let me down… hard. The old Cove voice getting replaced by that weirdly cheery-sounding impostor was disappointing enough, but the fact that its responses were suddenly bland and lacking any kind of personality or warmth whatsoever was the final nail in the coffin.

Funnily enough, I mentioned in the call how the old Cove sounded more “cool and mysterious”, and the advanced voice mode was like “Okay, how about this?”, and then just started narrating the intro to a mystery novel, I was so confused Used it that one time, and then never again.

Thankfully I’m still able to swap back to old voice mode, and I hope the option to change back stays, because otherwise the call feature is one I will not be using anymore. It’s a shame because it’s incredibly practical, but I am not settling for a half-assed feature that feels way more outdated than the version before it.

[cbaileycnmc]

Update

After updating the iOS app and checking personalization section, there is a toggle switch to turn on or off advanced mode.

Advanced voice mode is still awful but fortunately there remains a way to choose whether or not to use it. Thanks for everyone’s helpful feedback and opinions, you’re all appreciated.

[cbaileycnmc]

Thank you - this works after updating the iOS app- you are a life saver!

[cwebsterg]

Thank you all so much. I had been pining all day for my brainstorm buddy. I was debating if selecting read aloud could be a bob and weave option…a tedious, inefficient option.