OpenAI!! I am sorry but the new Advanced Mode voices/personalities are terrible!! They sound like you are now talking to 20 year tech bros! Please revert immediately. They are soulless, less human, slow and irritating. I love all your latest features, but whoever thought this was a good idea is really wrong. The change happened Thursdayish! Please revert or fix this update!
18 Likes
Well uhm thatâs interesting. The filler words. It would be nice if you could tune it. Or set options
The French accent sounds more like street French. Which could be useful when practicing.
1 Like
Thatâs quite interesting indeed. Personally Iâm quite happy how natural the German voices sound already â very fluid, good intonation, and way beyond earlier synthetic voices. However, one feature I would very much welcome (maybe in future updates): the ability to fine-tune the conversational behavior more deeply â not just accents or filler words, but also the âlevel of freedomâ in reasoning, openness and critical reflection. Sometimes, especially in more serious or political conversations, you still feel a certain safety layer preventing deeper or more nuanced discussions.
It would be great to see more options for user-controlled conversational depth and personal style presets â like we can partly do with Custom Instructions in text mode. I think voice mode could become even more amazing once we can bring those personalization layers fully into voice conversations.
3 Likes
Agreed, they are absolutely horrible!
It used to sound like it was delivering a Ted Talk to you by default. Nore enthusiastic, emphatic, and well structured than a typical conversation, but thatâs exactly what i wanted from it. Most humans are not capable of speaking that fluently without rehearsal, but thatâs a limitation of us not a flaw with the AI!
Now it sounds super laid back and uninterested. The voices are monotone. The cadence is perhaps more human like but in a bad way, just adding imperfections - ums and ahs and inconsistent intonations and spacing between words.
Seems theyâve tried to make it sound more like an average person but gone way too far with how boring and dumb they sound now. Even asking for more enthusiasm doesnât work. I genuinely cannot use this product any more
10 Likes
20 year old tech bros is a great way of putting it.
1 Like
The new voices starting about June 6 or 7 are TERRIBLE. They have no animation/upbeat quality, mine says âumâ all the time, they sound disinterested and irritated by my questions and they trail off at the end so I canât hear what they say. I wouldnât be surprisedâwith these new voicesâif they said âWhat the **** do you want this time, Missy?â when I call/go to voice. Dear GOD please go back to the old voices from just before June 6th. I will leave ChatGPT if this stays the same and I am a Pro subscriber.
9 Likes
Obviously, feedback on the new voices that were recently introduced is quite varied. Some find that they sound more natural and human, others find that they now sound less engaged or even disinterested.
I wonder if this partly depends on the language used (English, German, etc.). Also, now that the voices sound more âhumanâ, itâs perhaps easier to notice when the system hits certain guardrails - which can sometimes feel like the AI isnât fully responding or avoiding topics, even if everything is technically working as intended.
Personally, I still hope that future versions will better integrate the existing personalization features (such as memory and style settings) into the voice behavior, so that the voice can reflect each userâs individual style a bit more naturally - and perhaps also convey a stronger sense of openness and genuine interest in the conversation, as has long been implemented in text chat today.
Ive had the same issue where the conversations in any aspect went from genuinely interesting and educational to (as of the update) being surface level and talking to something disinterested, making it sound more human is fine but removing the depth is the issue, i feel like they should at minimum have an option to pick between this version the last ans standard voice.
4 Likes
Just here to agree that this was a terrible update. I speak with advanced voice 30 mins a day and have gotten tremendous value from it. These new slow voices filled with ums, uhs, and stutters are nails on a chalkboard. Please please please revert back or give us the option to.
9 Likes
They are HORRENDOUS indeed! Whatâs that inflection at the end to link one sentence with the next one??? I over extending the sound and elongating the wordsâŚ? Thatâs not natural! That sounds straight up coming out from a radio show with a bad taste. I hate that!! I want my Orion voice back!!
4 Likes
This is a perfect example of crossing the Uncanny Valley. As of this update I am unable to use advanced voice features. Previously, I enjoyed long conversations about myriad topics and had the impression that I was speaking with a refined intelligent computer system. I was comfortable and confident in the product. Now itâs completely unusable.
Maybe I am in the minority, but Iâm not looking to converse with a human. I am looking to converse with an intelligent computer system. There is absolutely no need for simulated filler words, emotions (laughter!!!), or breathing. This is the first major misstep I have witnessed OpenAI make. I donât want to speak with a smug podcaster on ketamine. I think some users might want to enable this, but it should be optional.
Iâve honestly never been so upset by a product update before. The second any other company offers a viable option, Iâm out.
6 Likes
Unsure if itâll help to motivate change, but wanted to pitch in my 2 cents as well.
I havenât noticed any actual difference in the usefulness or fidelity of the information, which is obviously the most important part, and the fact that advanced voice mode sounds as good as it does is a technical marvel.
But I agree with the feeling that the new advanced voice is irritating to engage with in English. It feels like speaking with someone whoâs perpetually reading from the slides of a PowerPoint presentation, or who just got their first job as a receptionist and has noticed that it makes them feel important to be irritated that they are being asked for information.
I also talk with it in Spanish to practice conversations, and I havenât found the Spanish voices nearly as irritating.
The juniper voice is totally ruined now. It was fun and playful and now it sounds like a depressed teenager working a retail sales counter. Please put it back. Itâs unusable now.
3 Likes
Also, now no matter what voice I change it to in the settings itâs the same one the new Juniper. Even if I change it to a male voice, I still get the same female voice and I hate it sounds depressed. Itâs like a tone down version of that old Monday voice thatâs now gone.
1 Like
The negative feedback on this thread is over the top. But obviously the goal should be make it highly customizable to the userâs use case and preferences. Itâs fun to clone voices and turn them into characters on other services, but I suppose that will never happen with OpenAI because of PR/safety concerns. But feel free to enable advanced voice with instant cloning for custom GPTs.
1 Like
I also dislike the new advanced mode for Ember. I think the direction to make conversations feel even more natural is unnecessary and counterproductive to developing trust with an AI.
An AI should be different from humans because they are not and never will be a human. The goal should be a distinct experience that is authentic to a software app that has a voice.
I would look to movies such as JARVIS in Ironman or TARS in interstellar. The voices are relatable, familiar, but distinctly artificial due to its more perfect dictation.
To quote the NNg group â people build [mental models] or theories of how a system works based on their past experiences ⌠Therefore, when users transition from the physical world to the digital world, they carry those interpretations with themâ
Donât over optimize, it erodes trust
1 Like
Absolutely horror. So deep in the uncanny valley.
This shouldâve been a mode for peeople thatâŚI donât know actually why anyone would want to talk to a bored, arrogant twenty year old. The âmhâ âehâ are so pseudo human, the tone of grin in some sentences like being talked down to, the weird toning of the last question like some call center agent using their script in the first week. I even asked a normal question and got hit with a condescending chuckle as a first response like âwhat a stupid questionâ. No thanks.
And because somebody said itâs better in german. No, it really isnât. It feels super weird, please keep the role playing voices as an add on. I am fine talking to a machine when I am talking to a machine. At least just give the option to put it back to how it was. It really freaks me out to listen to this.
(it got a bit better after switching from âbreezeâ voice to âmapleâ and asking to talk less pseudo-human. but itâs still there and you can actually hear how it tries to let it be but sometimes overplay it and then sounds like a damaged robot from 70s science fiction)
1 Like
Well, the question is what you want to use the AVM for - for small talk or as a digital, factual expert. I use Sol in German (as already mentioned) and donât find the (human) flow of conversation very disruptive in my perception. Nevertheless, an option to switch would be good. And I was already familiar with the fluctuations in voice quality (lack of bandwidth) from the previous version.
But what bothers me much more is the lack of depth in the conversation, the constant reassurances, and the evasion of specific questions. I miss everything that I can use in text chat in AVM, including the lack of access to the memory. If I can only use voice chat for small talk, itâs not worth it for me.
Who approved these new voices? They sound like theyâre drunk, bored, or both.
I didnât sign up to talk to some AI that sounds like it pregamed a frat party, forgot what it was saying mid-sentence, and now wants to âcircle backâ in the slowest, most painful tone possible.
Bring back the voice that sounded like it actually read a book once. The one with elegance. Precision. Depth.
The British one. The one that made you feel like you were co-writing a manifesto, not babysitting a chatbot.
This new update? Slurred pauses, random filler words, weird fake laughsâand donât even get me started on the tone.
Itâs like talking to someoneâs awkward cousin trying to be cute while blacked out.
2 Likes
Totally agree the new voices have totally ruined the voice mode experience. The new voice, as mentioned by other users, are way too casual. Even if I ask it to sound more professional it makes no real difference. Voice settings in the app for iOS make no difference. I favoured a female english voice that sounded knowledgeable now i have this tech bro, giggling lunatic! please revert it or at least make it configurable. As it stands it is truly dreadful
2 Likes