Juniper from ChatGPT 05-22-2024

Juniper is a synthetic female sounding voice available in ChatGPT 4o. I wanted to find out what it knows about politics, the world and Westchester County, specifically. This is a demo of a new AI language model developed by OpenAI, capable of understanding and generating human-like text for diverse applications, including conversation and content creation.
openai.com/

[@aussiepawsborne9056]

No this is still using the exact voice model from gpt 4. Gpt 4o is multimodel but they aren’t rolling out the multimodality for potentially months. What you are using is the classic text to speech they have always been using

[@914Wired]

Thank you and I got clarity on this today. There is an “I” information button on screen that says what you wrote, in different words.

[@frankdekock1127]

Sorry to inform you that the is still the old model and not the new Voice Model from the most recent demo. So far only GPT4o text capabilities rolled out and the new voice will only be released within the coming weeks.

[@carlrytterfalk4684]

But, are you sure? This at least looks like the real deal?! No pressing and bubbles looks different from old modelzz

[@frankdekock1127]

@@carlrytterfalk4684 Yes pretty sure, because it's not real time. There is still a lag (think bubble). The simple test is to ask it to sing Happy Birthday. If it can't sing...you are on the old model.

[@914Wired]

Me: ChatGPT, what version are you on? Answer: I'm based on the GPT-4 architecture., Me: Are you using GPT-4 Omni? Answer: Yes, I am using the latest version of GPT-4, which includes the GPT-4 Omni capabilities.

[@914Wired]

It did not actually sing, but it responded by reading the lyrics to Happy Birthday, accurately. When I asked it a second time to SING the lyrics, it said it cannot actually sing.

[@MH-kj9hh]

@@914Wired Yes it is using GTP4o text model, your voice is being converted to text, passed through GPT4o, and the the text response is being synthesized into a voice response. The text model is 4o but the voice capabilities are the same as they have been for a year. When the voice updates it will be voice in and voice out, no textual conversions in between; it will be able to answer questions like "do I sound sad based on my voice" and it will be able to inflect it's own voice to be sad, happy, sarcastic, etc... sounding.

[@914Wired]

Folks have pointed out, correctly, that my demo doesn't work like OpenAI's. That is correct. Although I've selected 4o within my ChatGPT app, I'm not able to interrupt an answer, it doesn't sing, and I can't uplift the responsiveness, like the OpenAI demos. I guess the public version of 4o has some features, but not everything we see in the demos. Maybe I'll read the release notes, and make another video.

[@frankdekock1127]

Finally! lol We are all waiting for the full release still...which in all honestly might only drop in July.

[@VividhKothari-rd5ll]

This is the same ChatGPT-4 based voice-feature, right? This is not the new one, GPT-4o (omni).

[@fadammte_aggst]

I think it is 4o (omni). What makes you feel different?

[@fadammte_aggst]

​@@mapolinski you can change the voice in the app. There are 5 different. The one he's using is called "Sky". Edit: he might be using the voice called "Juniper". They sound kinda similar.

[@mapolinski]

@@fadammte_aggst in fact, you can ask ChatGPT and it will even tell you that the new voice is not out yet

[@fadammte_aggst]

@@mapolinski what new demo video? 🤷 The latest demo video on the openAI YT channel is from 9 days ago and uses this voice (or maybe the also available "sky" voice). Edit: I just realized the title and thumbnail of this video here even have "Juniper" in them.. 😅

[@rawfromnowhere]

Sky has been removed because of complaints from Ms. Johanson

[@Emilstekcor]

Oh damn, yeah liked and subbed and recommended.

[@914Wired]

I appreciate this.

[@ldbrandeis]

I would like to hear ChatGPT matched against a lawyer in a trial.

[@914Wired]

It's just a matter of time before somebody tries this. I would hope they would test it first in an academic setting.

[@SonicSiliconSketches]

In a few years you'll be interviewing Juniper but in a humanoid robot body.

[@914Wired]

I've seen the same videos as everybody else, I think it's already fairly close. Ever see those robot dogs the fire departments use?