"Humanizing" Second Life

That's just a pretty way of phrasing an idea roundup of mine.

1) Avatar Text-to-Speech

For some time, I've wondered why this hasn't been done. Sure, we have resident attempts, but a backend system is really what's needed. Utilizing a program like Oddcast's SitePal, text could very easily be converted at the client side into speech. Imagine that for a moment.

2) LSL addition: llSpeech(string msg, float volume)

LSL equivalent of using such an interface.

3) LSL addition: animation(string recent)

Leaving the speech segue for a moment, we could also use a new event handler - animation. I know several people use animation overriders, especially when a new animation comes online - something like this would be perfect. Not only that, but it would allow us to better synch sounds to animations - something that's sorely needed.

Thoughts?

NOOOOOOO!!!

i don't want my avatar to sound like Stephen Hawking after he was kicked in the nuts

Heh. Naturally, the idea is to allow users to "switch it off" too.

Funny, I've always assumed that that is *exactly* what you sound like!

Personally, I would love to be able to switch between the voices of GIR, Johnny Number Five, Ro-man, and Ned!

pffffffffft.

Anyway, this 'humanizing' second life thing is speciest! You're assuming that we're all human here. Bloody humans... don't think anything exists outside their own little corner of the solar system.



(btw is speciest a word? Is it xenoist? Is that a word even?)

i was thinking the same thing after reading the thread on the animated lip sync experiment. i've not played/demo'd HL2, but the auto-lip synch in that game is what i really want to see (in everything, not just SL). when the Valve guys mentioned in their forums what modders could do (just add text), and also indicated that a multiplayer game could also have lip-sync driven by players using microphones, the interaction possibilities seemed very cool. (note: i don't know how much of that turned out to be hype before the game's release, so maybe it didn't turn out so great).

anyway, moving in this direction would seem the way to go imo to really make use of the new video capabilities. more machinima!

I think we need to stop proposing more work for SL and start coming up with ideas on how they can let us do the heavy lifting bandwidth / cpu wise.

A better idea would be a way to stream sound from our avatars, with a rolloff in sound volume as distance increases.

Then we can take this and write text-to-speach servers and what not.

Dunno, really. A runner to the idea, which I left out for precisely this reason, was horizontal "plugin" modification for something like this. It was left out precisely because, yeah - LL has a lot of work to do.

Naturally this is kinda long-term thinking - post 1.6 fix, post Havok 2, post even that new tree software Pathfinder pointed out.

As for us doing the "Heavy lifting," I can't see why not. With the exception of the raw soundclips being a bit space-intensive, the Sitepal client is fairly lightweight and low-calculation. And flash-based - meaning we could probably use it already if not for the evil lack of flash interactivity.


As for your little tangent, neither are words. The closest that comes to mind is philanthropy. There are several terms meaning quite the reverse, depending on your perspective. Regardless, can we stay on topic please?

I call it "humanization" simply because we generally equate present-day patterns of speech and body language in this manner - not to offend people who think differently.

I have some serious issues with your language, here... "Humanizing"? Not all humans are capable of or fluent in speech. And not everyone on the other ends of the computers can hear it. I enjoy the text. Call up your friends on the phone if you want to chat with them.

Yes imagine that...imagine 20 or more people in a room talking and not being able to distinguish anything because everyone is 'talking' at once. Text is just fine for me.

Having the regular typed text read aloud at the client side is an excellent idea, especially for those like me whose vision is failing with age in the normal presbyopic manner.

Modern text to speech systems can sound quite real and pleasant, with good tonal quality and a surprisingly good job of tempo and inflection. The "sounds like Stephen Hawkings" comment is like totally 20th century, man...

With not only an option to turn the whole thing off (which would be the default setting, of course) but also the ability to mute TTS on an avatar to avatar basic one would be able to more easily pay attention to the people with something worth saying without totally ignoring others by muting their text chat entirely.

Being able to assign specific voices to different avatars could lead to some amusing times, with the big tough looking jerks given teeny pipskeak voices and other such sillyness.

And for those who like cybernetic sex, the use of TTS script commands would make possible simulated sex objects mixing sounds clips with TTS like this:

llPlaySound( "You turn me on..." , 1);
llT2S("Philip J. Fry";

Plus those who operate multiple avatars at once could have one client read the chat text aloud while reading the other.

Suezanne, I was being deliberately obtuse as usual Yes, these days TTS can be pretty darned good - even for free!

My own alicebot is hooked up to a Lernout & Hauspie TruVoice engine - which is free and can be downloaded from M$$oft amongst other places, and I've played around with the AT&T Natural Voices SDK which is very cool indeed when tweaked right. You can play with a limited demo of it on the web here. Bell have a web demo of their system here too, including some odd voices such as 'coffee drinker' and 'gnat' The java version lets you tweak away at the voices and come up with your own.

There are other engines of course, some free, some trial. Try here for some demos and downloads.

Right on! I use Bell Labs online natural speech demo to record vocals for my music. Then I just load it into CEP and start twisting it. Sounds very good, plus it's fun to have to creatively spell words phonetically to get the right pronunciation.

SL forums need a spell checker

BTW: I think Genist might be best. That way we can remove all species not IMMEDIATELY connected to humans (including those cute fuzzy chimpanzees!) which, since every other Homo species is dead, leaves only Homo sapien!

Some versions of windows (XP for example) come with bundled text-to-speech. It sounds okay. As I recall, LL's goal in that area is to get something together that would let us talk to each other via microphone, and do masking, in order to A) allow some anonymity and B) let people make their voice sound a little different, or like the opposite sex, or like some kind of machine, or heavily distorted like when they interview a whistleblower on 20/20.

Sadly, I've had experience with this.

PSO for the XBox had TTS with a converter.

The problem was everyone sounded like 12 year olds, regardless of the filter.

As far as I know, realtime voice manipulation just hasn't reached very great heights yet.

Try the Oddcast link I left in the initial post on this thread. You'd be surprised.

I personally really dislike this idea. I rarely spend more than 25% of my attention to SL conversations (unless I'm having a very complicated conversation about a very consuming issue, technical or otherwise) and hate to be "odd man out" all the time because I refuse to focus 100% of my attention talking into a microphone.

Proposals that require almost no development time are the best. Here's one.

Simply write the chat text that already appears on our screens to files, one per chat/IM/etc channel.

Then anyone who wants sound output can simply point their speech synths at these files, and get live speech out. Job done! I do this all the time on IRC, I've had it working like that for years.

The simplest solutions are the best, and this one works great, consumes effectively no resources (certainly no network bandwidth, because the text data is already coming down to us), and could be coded in just a few minutes.

Then from that easy beginning you can figure out more complex schemes, when there's time.

Morgaine gets my daily brilliancy reward.

Indeed.

Personally, I'd consider writing it myself with a little XMLRPC... but nope, replies from Second Life only.

As for "more advanced features..." oh, dunno. Client-side lip synch? Better to go after the root idea first.

Oooh! Oooh! Idea!

In the same way that we have sliders for avatar customization, there could be sliders to customize either/both the text-to-speach voice or voice mask, i.e. pitch, echo, animal or machinelike effects...

Of course, all this gets immensely more complicated now that LL has made the commitment to multilingual text support. I've seen some of these TTS tools (the Mac ships with some), and they're not that bad... but they generally aren't available for Chinese, for example. Some are, but not the free ones, at least, not that I've seen so far. And speech-to-text for non-European languages is even harder to find.

On the other hand, if there *were* good TTS and STT converters for other languages, the prospects of creating good language learning environments in SL would be fabulous.

I'm inclined to agree that being able to automatically write out chat/IM logs for TTS processing would be a Good Thing (heck, it would be good anyway, even without TTS). Getting the STT input INTO SL, on the other hand, may be a bit longer in coming.

neko

The open-source Festival is quite excellent in its generated speech quality. Some of its better voices sound more human than some people I know ...

Bumping this. I would find text-to-speech very useful, particularly if I could direct it at a particular avatar. It would require a permissions request I suppose.

Lip-sync-ing by way of a mic plugged into the computer ?
That way your ava will sound like, you, wouldn't even need text to speech at all, just a mic & your voice would come for your ava, then you'd know for sure who's talking as how many folks sound alike ?
Then LL can get rid of text alltogther as folks would just be 'talking' well, IM would still have to be in place for long distance conversation I suppose, but think of how cool it'd be for your own voice to be comeing out of your ava otherwise ?

Woulfe
* No Longer On FurNation Sim, If You Want To Know Why Just Ask Me In An IM *