zlacker

Which one are we saying sounds like Johansson? I'm talking about the TTS voice in the app, is everyone else talking about the multimodal voice from the 4o demos?

Also, whether they *intended* to replicate Her and whether they *did* in the end are very different.

>>stavro+(OP)
This one: https://youtu.be/vgYi3Wr7v_g?feature=shared&t=22

compare it to: https://youtu.be/GV01B5kVsC0?feature=shared&t=125

>>m_ke+m
Well, I thought it will be similar, but at least with how sky voice sounds through the phone speakers, I can hardly find any resemblance.

>>m_ke+m
I keep reading in the media that Sky was introduced as part of ChatGPT-4o, but that's incorrect. Sky's been around since they introduced the mobile iOS app.

While Sky's voice shares similar traits to SJ, it sounds different enough that I was never confused as to whether it was actually SJ or not.

>>maroon+e2
I don’t think you understand. 4o introduces a new multimodal Sky replacing the old one. They have only released clips of the new voices. It’s never been in the iOS app. The one you refer to is the old voice model. If you listen to the linked video above it’s very obviously not the same voice (I use Sky on iOS btw)

To be honest the new sky is obnoxious and overly emotive. I’m not trying to flirt with my phone.

>>m_ke+m
I am of two minds here, regardless of the "closeness" there is a whole field of comedy that does impressions of others. That is what is so difficult about the AI discussion. Clearly, there are plenty of humans who can mimic other humans visually, in prose for writing, in voice and mannerisms etc.

Leaving the IP issue aside, they could clearly have hired a voice actor to closely resemble Johansson maybe without additional tweaks to the voice in post processing. If they did do that, I am not totally sure what position to take on the matter

>>m_ke+m
Those don't sound anything alike, except being two female voices. Sky is clearly a bit lower and with a lot more vocal fry.

>>m_ke+m
Thank you for providing a nice side-by-side. This makes it clear to me the voices are not very similar at all. If Johansson had agreed, I have to imagine they would've been able to make a much closer (and less annoying!) voice.

>>m_ke+m
OK, I watched this expecting to be convinced.

I think they might have mimicked the style. The voice, though, is not even close. If I heard both voices in a conversation, I would have thought 2 different people were talking.

>>skg_RS+J8
The important thing is that they never said it was Johansson. They were not pretending to be her. They are not imitating her likeness whatsoever.

>>m_ke+m
Are you using this as an argument about how similar they are? The voice sounds distinctly different, no problem discerning between the two.

>>m_ke+m
Holy Crappyness Batman! The OpenAI clip is so bad. Homeboy keeps stepping on "her" lines. So from this I come away with he's just a rude asshat that doesn't know how to socially interact with people, she's just too damn chatty and doesn't know when to shut up, or maybe it was just really bad editing? Either way, it's not an intriguing promo to me in the least.

>>Freeby+qb
Some employees were definitely thinking of Scarlett Johansson, even ignoring the reference to the film "Her":

https://x.com/karpathy/status/1790373216537502106

>>Beetle+bb
Without commenting on the debate at large, it’s a bit funny to read this comment.

I mean voice cloning a year or two ago was basically science fiction, now we’re talking about voices being distinguishable as proof it’s not cloned, sourced, or based on someone.

FWIW I also thought it was supposed to be the her/sj voice for a long time, until I heard them side by side. Not sure where to stand on the issue, so I’m glad I’m on the sidelines :)

>>flumpc+dk
Karpathy doesn't work for OpenAI anymore tho.

>>Beetle+bb
I agree they don't sound the same. But, since it's a subjective test, OpenAI was pretty Twitter-foolish to push the "Her" angle after being explicitly rejected by SJ. It's just inviting controversy.

>>dilap+0a
The cadence and speed in Her is much too fast for any mass customer product

>>Beetle+bb
Truthfully, you can no longer trust yourself (whichever side you're on in this debate). We're all now primed and we'll pick up any distinguishing characteristics. You'd have to listen to them in a blind test and do this with several clips that do not reveal which ones are OpenAI and which are from a movie or something else that spoils it.

And I wouldn't put the metric at 50/50, needs to be indistinguishable. It would be a reasonable amount where it sounds __like__, which could be identifying the chatbot 100% of the time! (e.g. what if I just had a roboticized version of a person's voice) Truth is that I can send you clips of the same person[0], tell you they're different people, and a good portion of people will be certain that these are different people (maybe __you're different__™, but that doesn't matter).

So use that as the litmus test in either way. Not if you think they are different, but rather "would a reasonable person think this is supposed to sound like ScarJo?" Not you, other people. Then, ask yourself if there was sufficient evidence that OpenAI either purposefully intended to clone her voice OR got so set in their ways (maybe after she declined, but had hyped themselves up) that they would have tricked themselves into only accepting a voice actor that ended up sounding similar. That last part is important because it shows how such a thing can happen without ever explicitly (and maybe even not recognizing themselves) stating such a requirement. Remember that us humans do a lot of subconscious processing (I have a whole other rant on people building AGI -- a field I'm in fwiw -- not spending enough time understanding their minds or the minds of animals).

Edit:

[0]I should add that there's a robustness issue here and is going to be a distinguishing factor for people determining if the voices are different. Without a doubt, those voices are "different" but the question is in what way. The same way someone's voice might change day to day? The difference similar to how someone sounds on the phone vs in person? Certainly the audio quality is different and if you're expecting a 1-to-1 match where we can plot waveforms perfectly, then no, you wouldn't ever be able to do this. But that's not a fair test

>>dylan6+Lf
I think the whole thing was scripted beforehand and approved by Sam Altman, of course.

>>onemor+HS
That doesn't really make it better because now a) it was a horrible script, b) the fact they didn't try to clean up the audio from "her" with anything more than a fade. If you told me this was just some intern making a video, then maybe, but now you've told me it was scripted just oh so makes it worse to me.

>>m_ke+m
The OpenAI one is recording the audio from a phone, where as the movie version is into a mic directly. They will sound different, but there are elements that are the same. Anyone using these to compare though and saying they don't hear the difference isn't comparing apples to apples.

However, the fact that there is a debate at all proves there should be more of an investigation done.

>>fnordp+W7
I've listened to the clips and yes, while 4o Sky is more emotive, it's just that - a more emotive Sky. All the elements that people are pointing to - the husky/raspiness - were present in the pre-4o Sky.