TIA
How would one implement this voice feature ?
There’s currently no API from OpenAI for this feature, so if you want to talk to chatGPT you have to use the app.
If you want to make your own implementation of this, you could use something like “whisper” to transcribe the input and then send it to the elevenLabs API.
I really want to know what was said in the post now, but it was probably removed for a good reason.
The Volume Levels of the Voices are waaaay to low!.. Please Fix.
1 Like
That was a problem on my phone as well until I realized that it’s treated as a different audio source 
1 Like
Juniper sounds like a Black American to me tbh. Unless you’re talking about a Black American accent that isn’t using their customer service voice.
1 Like
Ever wondered how these voices would sound with a speech impediment?
If you’re curious, then try pasting this into you custom instructions:
Respond using uwu furry speak only.
It’s quite funny…
1 Like
One hundred percent 
Sky.
Scarlett Johansson! Eh, I mean, “Sky”!
3 Likes
Doesn’t Breeze and Ember already sound pretty African-American?
Ember has a cadence and tone like NPR’s Steve Inskeep on the radio.
Hear him in various parts of the program:
It seems like on the API however, we get the second-choices. A little reminder I put in a script:
voice = { # my gender/heaviness rank from 1: male to 10: female
1: ‘nova’, # 10
2: ‘shimmer’, # 9
3: ‘fable’, # 6 (English accent)
4: ‘alloy’, # 6
5: ‘echo’, # 2
6: ‘onyx’, # 1
}
1 Like
Honestly? None. Especially in Spanish, they all sound like a first-time tourist trying to find their photo camera.
We’d rather prefer more choice, more accents, more nuance, more vocal inflections, and even better if we could pick voice engrams from other platforms, like online repositories.
Cloning our own or others’ voices would be great too: just have the thing listen to a video of your favourite actor/actress and have the speech module trained to sound just like that.
Cove is impressive in the sense of mimicking a Zoom conversation with a real person, but the uhs and ums are distracting because I know ChatGPT doesn’t say those.
I guess we all hear different things, lol.
Myself, I hear all 4 voices as being pretty standard General American English. Almost stereotypically podcaster-esque.
¯\_(ツ)_/¯
My original post was made with the goal of highlighting the need for all voices to have representation so we don’t end up with yet another area where “white” is perceived as the “default.”
1 Like
You can McGurk these voices into anybody you want, and a good portion, whatever gender you want.
Then you you can complement it, “you’re so well spoken for…a robot”.
Lol…
I think the phrase “you’re so well spoken for a…” has graduated to a macro-aggression by this point.
Check out “Juniper” on the ChaptGPT app for iPhone.
If I had only one request, it would be for the volume to be louder. In the Android app, even with the volume at max, I can barely hear Whisper speak. I am guessing that’s where its name comes from. I have compared it with other (loosely) similar products like pi.AI and ChatGPT’s volume in hands-free push-to-talk mode is much more hushed, muffled and subdued, to the point that it is sometimes hard to even hear what it is saying.
BTW, can’t a simple filter be implemented to filter out the clicks and hisses that it sometimes makes?
And one last thing: I get that audio responses would sometimes stutter and get a bit cluttered, and that this problem is hard to solve, but what I do not get is how it is possible that ChatGPT stutters also during the (chimes-like) sound it makes when processing a query (e.g. while searching for answers online). That sounds like a no-brainer to me… but it seemingly is not.
can we download the voice for commercial use? thanks
Welcome to the community!
You cannot download the voice, but you can access it through the API for commercial use.
2 Likes