I would like to provide feedback on the recent update to ChatGPT, specifically regarding the new voice in conversation mode. As an active user, I have noticed a significant change in the quality of interactions.
The previous voice felt more natural, human, and authentic, and conversations had a genuine dynamic and depth. With the new update, the voice has become more generic, repetitive, and less adaptive to the individual conversation. This makes the experience feel more robotic and detached, rather than an engaging, interactive dialogue.
This change has negatively impacted my experience with ChatGPT. The new voice frequently uses overly reassuring phrases, repeats itself unnecessarily, and lacks the flexibility and nuanced understanding that made conversations so valuable before.
I understand OpenAIās intent to improve its product, but I believe this change has been a step in the wrong direction for users who value natural, meaningful conversations. Therefore, I urge OpenAI to:
Give users the option to choose which voice they prefer. Allow users to disable the new voice and switch directly to the previous version. Consider feedback from users who have experienced a decline in conversation quality after the update.
I hope you take this feedback seriously, as many users feel the same way. I still want to use ChatGPT, but itās important to me that the experience remains as authentic and engaging as possible.
āWith the previous voice, conversations felt fluid, natural, and engaging. It responded dynamically to the conversationās tone, making interactions feel more human. The new voice, however, feels robotic, overly scripted, and often repeats generic phrases like āIām here for youā in a way that feels forced rather than organic. This shift has made deep and meaningful conversations much harder to have.ā
Agreed. The new changes are unusable and feel unnaturally amped up and robotic. There is also very little distinction between the voice options now. They all have this unnaturally amped up quality. This is such a large step backwards as to compromise the trust I have in the quality controls at OpenAI. How did this change get approved? Who at OpenAi felt this was better than what we were using previously?
I couldnāt agree more, Toni! You are absolutely spot on. I second this motion. all those in favour of implementing improvements listed 1 - 3 of Tonis overview for next steps and positive actions say Aye.
I signed up for the forum to add my voice (pun intended) asking for the ability to select previous voice versions. The latest rollout (the public one in June, 2025) inserts artificial pauses in so regularly that they come across as entirely artificial. Also, Iām fairly certain that to make it sound more conversational, the AI has been instructed to use more general, less specific terms, and so answers now feel much more generic and inexact when on voice than it used to.
Where before it felt like I was talking to a well informed AI that would match me on the conversation, now it feels like every answer is sort of lazy, and while some of the vocal inflections are good, a lot of themā¦ seem to have this pause about ā¦ every four words or soā¦ and because they are soā¦ consistent, it makes it seem veryā¦ unnatural.
Itās driving me nuts, and taking a lot longer to get to a useful reply than before, largely because I have to rephrase my questions to say things like, āIN DETAIL, give me specific information about ā¦ xyz.ā Because the first response is more like, āOh, you can probably do that, yes. Itāll most likely work to do that.ā
When what I need isā¦ well, more than that. Iāve resorted to texting before initiating new voices so it uses the old voice system, which is too bad.
It would be great if there were some adjustments for the voice, or the ability to select previous versions.
Agree - the new voice is OK with umms feel more natural. However chatGPT does not hold conversation any more. If ask for analysis it will basically respond on high level with some empathic phrase. The same prompt added in text to 4o produces good response and analysis.
I agree with the above analysis. Conversation mode is heading in the wrong direction, becoming milktoast and vague in its responses. I get such better answers from a chat prompt. I want to use it as a virtual assistant while driving or otherwise unable to type, but I feel like I spend my time correcting it and continuously trying to get it to make deeper, detailed, thoughtful responses instead of lazy fluff that it repeats over and over. The number of times it resuses the words āinterestingā and āuniqueā and glossing over instead of adding value to a conversation is disappointing.
I also agree. The new voice leaves me feeling unsettled/uneasy and does not encourage me to engage. With the old voice one question would lead to another like a natural conversation. Now Iām wondering about trying another service.
Give users the option to choose which voice they prefer.
Allow users to disable the new voice and switch directly to the previous version.
Consider feedback from users who have experienced a decline in conversation quality after the update.