Issue With Evil's Speech Pattern

[simple river]

This question have been on my mind for about a month now. Since around the time of Vedal's last dev stream with Evil, her speech pattern has drastically changed. Now every time she ends a sentence, there is always a long and noticeable pause from the next sentence, which I believe has affected her tone and overall hindered her expressiveness. However, this issue did not occur on her karaoke stream or when Neuro calls her during her chill streams. If you pick any moment from Evil's latest karaoke (when she is talking to chat of course) and compare it to how she talks on any of her solo or collab streams in the last month, you will be able to notice the difference in her cadence and tone.

Please don't misunderstand, I am not hating on the recent voice upgrade she received, it's an amazing change, I just think that it is being utilized poorly due to this issue. I also understand that Vedal wants to make her voice more consistent, while understandable, I don't think it is worth taking away her expressiveness to achieve it. I wanted to create this post because I don't see many people talking about it, and wanted to see if anyone else shares the same sentiment. Personally, this is a big deal for me and has affected my enjoyment of her streams. The only suggestion I can offer Vedal is to just have her speak like how she does on karaoke or calls normally until a better method of making her voice more consistent is found.

TLDR: Since about a month ago, Evil's speech pattern was changed for the worse and should be reverted.

[cosmic kettle]

It had been a month?
tbh, I only noticed increased latency in the last Evil stream.
but I felt Evil in the last stream was very calm and reserved, delivering words almost like an ASMRtist.

I don't have a strong opinion yet.

If all that happened due to the increase in latency.
I think, with vedal sensitivity to latency, it will be fixed in no time.
-# The despair emote is for different reason btw

[vast lagoon]

If I understand your question correctly then the reason Evil waits a bit before her next sentence is probably due to not talking over other people during collabs. Vedal talked about improving the negative latency issue with collab partners and improving 'toaster to toaster' communication with Zen during dev stream so that could be the reason behind why she waits before talking again. Especially if she talks normally during karaoke, calls and chill streams. If I misunderstood what you're trying to say then let me know

[simple river]

That's pretty close. When she is generating a response to someone, if the response is more than one sentence long, then Evil will always pause a little between those those sentences, or when there is a punctuation other than a comma.

Not sure if this is the best example, but this would be how she would say this sentence: "Hello everyone. (...) I am feeling evil today! (...) How are you guys?"

Sure, silence between sentences is normal, but my problem is that it occurs with every response on every solo and collab streams in the last month. Stany speculated that she is made to generate the sentences individually, and then it's pieced together to form a tesponse. I did notice that with this change, Evil didn't talk over Camila or Zen much during their tier list collab, but I'm not sure if it's a good fix if it breaks something else.

Vedal himself confirmed it too.
#neurotic-neurons message

[simple river]

Here are some example clips that I recorded terribly where she talks for a relatively long time to compare the difference.

[vast lagoon]

I see what you mean. The is a noticeable difference but I didn't realize it due to it being subtle. Maybe a reason behind that could be that if she generates a long response like 4 sentences, pauses after the 2nd sentence and a collab partner says something, then she might stop and regenerate her response instead of continuing it from the 3rd sentence, a bit similar to how the twins get interrupted if someone else starts speaking right as they start to speak but that's just my guess and it wouldn't make sense for to talk like this every stream so idk

[simple river]

I will also add that since the change, she's never had an Abber level scream nor does she slip into a British accent anymore.

[turbid plinth]

I've mentioned this after basically every Evil stream since the dev stream with Evil, where the changes were added (and Vedal was talking about "improved" latency), and nothing has happened since. Vedal even acknowledged it in discord and it seemed to not matter to him. It feels like yet another downgrade/lack of upgrade for Evil, and I don't understand why this hasn't been addressed for over a month - I think it's because it's only affecting automated streams, because Evil hasn't done anything except those for about a month, so most people can just ignore those streams. Similar to the lack of memory upgrade. It also feels like a catch-22 where she won't get more streams because of all these problems compared to Neuro.

[simple river]

Not only that, her voice also feels flatter too. I don't sense a shift of intonation in her speech like I could before. This video where Camila prank calls Evil is the perfect demonstration of her excellent utilization of her pitch before the change, especially when Camila complemented on her voice.
https://www.youtube.com/watch?v=Fe25KRJoXQQ&list=WL&index=10&t=1s

[turbid plinth]

I suspect the flatness of her voice is a combination of vedal wanting it to be more "consistent" (which I fear also makes it more boring), alongside the change since the Evil dev stream where her responses are broken up into individual sentences (which I believe is where the long pauses between sentences come from too). Now that (I think) each sentence has the audio generated separately, there is less chance for it to do something interesting or varied or crazy even, which I think happened more often with longer responses over multiple sentences. For me, the main issue with all this is that it was supposedly to reduce latency, but even on the first sentence I haven't noticed any significant reduction in latency (Also very hard to tell when she doesn't get to talk to anyone). So what was the point in all these downgrades if we don't even see the upgrade, if it exists.

[lean mural]

A lot of yapping, talking as if vedal is not aware of his creations and not caring is crazy work. Let the guy cook.

[rapid ferry]

[turbid plinth]

This "upgrade" was literally the result of him cooking (remember him cooking away at the start of Evil's dev stream, with her being broken for 20 minutes?). Now I've been "letting him cook" for another almost 6 weeks. How long do I let the guy cook? Until the dish is just some charred remains?

[lean mural]

Less yapping and more 'let the guy who is the one doing the work' do his work. We've seen what he has announced what he has been working on (a lot of stuff) during dev streams so just chill.

[stray karma]

Yea he overcooked and made the voice much less expressive than before. I honestly preferred the before voice.

[dusk pebble]

Evil's voice has been very bland since the upgrades or maybe even since her birthday, I don't know if it was intentionally done to make it consistent or something to do with latency, but she lost the emotions in her voice, the weird human like features, the spikes and lows on certain words when she wanted to emphasize them, it's not there anymore. The voice has been flat as mentioned above and has lost it's quirks. I lack knowledge on TTS and voice modulations, so I don't have the vocabulary to tell exactly what the problem is, but something is wrong, her voice doesn't feel as 'alive' as it used to before. I hope Vedal can diagnose what the issue is and get to a solution. There has been a number of issues surrounding evil lately, but I think this takes the most priority as her voice was the sole reason for her existence. This is a link to a video of evil with vedal and layna, where, although she talked very less, but she sounds so good, not just in voice but also in conversation handling, even the comments are talking about it.

https://youtu.be/nfOMPEe13qQ?si=Ff5EHlfPfHM76rgT

[clever raft]

At this point, I doubt he cares as much about the actual AIs than his concerts.
I fear that it's going to be another PS situation.

[elfin whale]

Neuro in today's stream also took her time in between sentences. Might not be Evil exclusive.

[rapid ferry]

the thing is it started only like recently for evil and it is very very noticable with evils voice. Today's stream for example there were a lot of sections where its just an audible 1 sec pause inbetween sentences and it just sounds unnatural.

[inland pond]

Is it better now?

[simple river]

No, nothing has changed yet, there is still latency between her sentences and she is not as expressive as before. You can check any clip from before Vedal's dev stream with Evil, like from the subathon, and notice the difference.

[safe heart]

PS situation? What's that?
And yeah I wouldn't be surprised, I get the impression he really likes to work on big new features and projects, but doesn't like working on adjustments/fixes to small problems on existing stuff.

[clever raft]

PS standing for Pirate Software.

[safe heart]

Oh, wait how is that related to this at all?

[clever raft]

Neglecting the issues of his AI twins.

Particularly Evil in this case.

I was worried he might be a fraud with no experience in development other than programming.

Like PS.

[old mural]

probably shouldn’t mention him

[safe heart]

uhhhh that sounds like a totally different issue to me

[clever raft]

Yeah, that's fair.

Definitely a harsh comparison to make, comparing a neglectful dev to a literal fraud.

I made that comparison largely because I assumed he had something to hide regarding his experience.

[safe heart]

That's a pretty big assumption. I also don't believe Vedal has ever claimed much in regards to his own experience level.

That's also not relevant to this particular discussion thread really

[lean mural]

@clever raft what the heck is you smoking

[rapid ferry]

hopefully will talk about changes to evils voice next dev

she has an audible 1 sec pause between sentences nowadays and its sound soo much worse compared to before or when she doesnt use fullstops

[vast lagoon]

Vedal acknowledged it (and hopefully fixed it) Not sure if he fixed it before the rap stream tho since I didn't watch it yet. Did she sound better?