Voice Support for the Linux Client

ESD should be run as a root daemon. I didn't know it could be run the user session.

I have an X86_64, but because I was having issues with multimedia stuff, I started using the 32-bit version of OpenSUSE. I haven't noticed any performance hit doing this.

Adamas

I just tried launching esd as root, and pulseaudio as normal user... and upgraded to 1.19 - is still not working!

I noticed improvements anyway - with pulseaudio alone, the audio is behaving strangely (i think it is a 48000/44100 issue) and this is solved now...

Also, I can see alsa in both capture and output in the voicechat device setup now, but i am not able to test the capture as I assume the test meter is still broken under linux!

in addition now I am alone in the Active Speakers window!!! I suspected a firewalling issue, but this is unlikely as i have not touched anything... and tthis happened before, when trying the various audio daemons combinations... so i will keep bashing my head against it... and i will post the results...

I GOT IT WORKING!!! I'm using a Logitech USB based headset in addition to my regular audio. Here's what I did.

VIVOX USES OPENAL TO USE AUDIO. It's configuration is in ~/.openalrc. This is my configuration, using a USB headset as the second audio source:

(define devices '(alsa native esd null))
(define alsa-out-device "hw:1,0"
(define alsa-in-device "plughw:1,0"

Yes, they use Lisp.

Second, use ESD to direct all SL sounds (but not Voice) to your regular sound system. You may also use ALSA, whatever works for you. Run esd in the background.

Then run SL... and connect... with voice!

It might be cool if, in the client or configuration, there were a way to specify the device to use for i/o manually like in a lot of other linux applications. Just a thought.

Drake, what kind of system do you have? I tried to configure OpenAL on my OpenSUSE 10.3 (the rc file is in /etc, not in my home folder) and when I configured it like that, it broke voice. I don't think I'm using OpenAL at all, or at least not that way.

I wonder if this is why LL is having such a fit with Linux voice. With this kind of difference, it may be necessary to post instructions distro by distro, rather than any kind of "one size fits all."

Adamas

here is where i get confused:
>In short, ALSA and ESD are daemons (system services)
For ALSA is not a deamon but part of the kernel (see http://en.wikipedia.org/wiki/Advanced_Linux_Sound_Architecture) i thought all was done speaking about ALSA has to be installed. Reading this thread again and again i see the insisting on that alsa deamon.
So there must be one, most probably a deamon named alsa accessing alsa on kernel.(that would confuse me to get those things apart even in german).
So where is this daemon to find for debianesque systems (like ubuntu)? (if you dont know
/path/to/filename on your system would help me google it)


Armin

Okay, a proper explination to all this ALSA/ESD/OpenAL, etc mess.

ALSA is the latest Linux sound system. It replaces OSS and OSS/Free (which is in the Kernel). ALSA just released 1.0.16, and the kernel has 1.0.15 of the driver package. Unlike OSS, you also have a set of libraries that'll let you max out your experience. Most programs using sound link to ALSA. You don't run ALSA like it's a server -- most of the time the /etc/init.d/alsa service just sets the initial mixer settings. Anything sound runs through ALSA.

ESD was one way to mix mulitple sounds together, and cache common sounds. It's a proper daemon, or server. It was written before ALSA's debut. On the up-side it allows mulitple programs to send audio data down the line and have 'em mixed. On the down side, it's dog slow -- by nature it delays audio about a second. There's many drop-in replacements, like PulseAudio... but the only thing that works for me is the origional ESD. ESD now interfaces with ALSA, which provides it's own real-time multimixer called dmix, and on certian cards like the SoundBlaster Live! and Audigy ES (aka emu10k1 cards) it'll use the hardware mixer.

OpenAL is short for Open Audio Layer. It's like OpenGL and SDL for audio. Initially independent, Creative Labs has rescued it and keeps it up to date.

Now what does this all mean for our SL client and voice?

SL itself uses the FMOD library, which can connect to ESD, OSS, and ALSA in that order. Since most of the sounds aren't really time sensitive, ESD is rather good. SL's Vivox voice system doesn't. It uses OpenAL, and runs as a seperate process. That means unless you have a hardware mixing sound card (an emu10k1) you'll have trouble trying to merge the two.

Which bring me to this point: Don't treat SL and Voice as one when they're not built as one. Treat them seperately, and give them a sound card each.

Yes, I'm saying go out and buy a decent USB headset (I like Logitech) for $25-$60, or get a dongled USB to two-port headset adapter. Then like I did above, route voice through it.

It's in my signature. Are you using a USB headset?

and which headset ? (given it is linux _and_ slvoice complient?
Guess even $50 is cheaper than messing around with loads of configs which worked for me with any sound/voip software without having to be actually touched for years.

Thanks, Drake for the "enlightenment" on Linux audio.

But are you saying that we can't do a simple software-mixed VoIP in linux, when it's easily done in Windows? I've always used my on-board sound cards sucessfully with VoIP!

Ok... I got voice working at the end... with pure ALSA...
1) replaced the integrated audio with an audigy 2
2) eanbled alsa in sl
3) I have this in ~/.openalrc:

(define devices '(alsa))
(define alsa-in-device "hw:0,0"

Only problem i have with SL and ALSA... the audio after a while stops... until I restart SL!
No messages are posted in console syslog when this happens.

Any clue?

I have HDA Intel and SL playing sound on ALSA, oss is stuttering badly, ESD not working.
With the suggested .openalrc voice is not working

2008-02-06T20:43:59Z INFO: LLVoiceClient::addCaptureDevice: Advanced Linux Sound Architecture (ALSA)
2008-02-06T20:43:59Z INFO: LLVoiceClient:etState: entering state stateConnectorStarting
terminate called after throwing an instance of 'std:ut_of_range'
what(): basic_string:ubstr
2008-02-06T20:43:59Z INFO: LLVoiceClient::clearRenderDevices: called
2008-02-06T20:43:59Z INFO: LLVoiceClient::addRenderDevice: Advanced Linux Sound Architecture (ALSA)


At the same time, SL stops playing sound until I restart it.
It is strange because i can use aplay or other players with alsa while SL is on, but voice seem to clash

I understood that SL should not be using ALSA, only Voice. Try disabling all SL sound with the following lines in the Secondlife script:

export LL_BAD_ESD=x
export LL_BAD_OSS=x
export LL_BAD_ALSA=x

That should tell you if VOICE and only voice works correctly with ALSA (close anyother programas that could use ALSA sounds also, IM, Players, etc).

If that works you can try running ESD and force SL for ESD like:
#export LL_BAD_ESD=x
export LL_BAD_OSS=x
export LL_BAD_ALSA=x

The ESD daemon has to be running... I can started as my user or as service. A run of "esd" should do it.

Everything seems to be working, I see no errors about LLVoice... I copied some of the output below.
I see my white dot, but noone else's. And only see my self in the Active Speakers list.
I have ALSA all free for the LLVoice.... SL without any sound (FFMOD didnt even load).

Shouldnt it work? Are any messages below relevant?
I'll be in the Voice-aholics Headquarters trying to test this.. and in the Linux Client Users group.

Anyone care to help us figure this out?

-----output
2008-02-06T21:39:24Z INFO: LLVoiceClient:essionStateChangeEvent: left session sip:confctl-314@bhr.vivox.com, name handle
2008-02-06T21:39:24Z INFO: LLVoiceClient::removeAllParticipants: called
2008-02-06T21:39:24Z INFO: LLVoiceClient::removeParticipant: participant "xLXxjXNifQ122c-yUrDCnBA==" (2d7c635c-d89f-435d-b673-ec94ac30a704) removed.
2008-02-06T21:39:24Z INFO: LLVoiceClient:essionStateChangeEvent: left session c1_m1000xLXxjXNifQ122c-yUrDCnBA==0in state stateRunning
2008-02-06T21:39:24Z INFO: LLVoiceClient:etState: entering state stateSessionTerminated
2008-02-06T21:39:24Z INFO: LLVoiceClient::notifyStatusObservers: STATUS_LEFT_CHANNEL, session URI sip:confctl-314@bhr.vivox.com
2008-02-06T21:39:24Z INFO: LLVoiceClient:etState: entering state stateNoChannel
2008-02-06T21:39:24Z INFO: LLVoiceClient:etState: entering state stateSessionCreate
2008-02-06T21:39:25Z INFO: LLVoiceClient:essionCreateSendMessage: requesting join: sip:confctl-314@bhr.vivox.com
2008-02-06T21:39:25Z INFO: LLVoiceClient::notifyStatusObservers: STATUS_JOINING, session URI sip:confctl-314@bhr.vivox.com
2008-02-06T21:39:25Z INFO: LLVoiceClient:etState: entering state stateJoiningSession
2008-02-06T21:39:28Z INFO: LLVoiceClient:essionCreateResponse: Session.Create response received (success), session handle is c1_m1000xLXxjXNifQ122c-yUrDCnBA==1
2008-02-06T21:39:29Z INFO: LLVoiceClient:essionStateChangeEvent: joined session sip:confctl-314@bhr.vivox.com, name handle
2008-02-06T21:39:29Z INFO: LLVoiceClient:etState: entering state stateSessionJoined
2008-02-06T21:39:29Z INFO: LLVoiceClient:articipantStateChangeEvent: state is 7
2008-02-06T21:39:29Z INFO: LLVoiceClient::addParticipant: participant "xLXxjXNifQ122c-yUrDCnBA==" added.
2008-02-06T21:39:29Z INFO: LLVoiceClient:articipantStateChangeEvent: added participant "xLXxjXNifQ122c-yUrDCnBA==" (2d7c635c-d89f-435d-b673-ec94ac30a704)
2008-02-06T21:39:29Z INFO: LLVoiceClient::notifyStatusObservers: STATUS_JOINED, session URI sip:confctl-314@bhr.vivox.com
2008-02-06T21:39:29Z INFO: LLVoiceClient:etState: entering state stateRunning
2008-02-06T21:39:29Z INFO: LLVoiceClient:endPositionalUpdate: Setting speaker volume to -11

Works fine in my Ubuntu 7.10 32bits. Still have problemas with Fedora8 and ArchLinux.
Could it depend in some libs that I dont have installed?

Now that I actually got to know where people were talking, I could make some more tests.
It seems that exactly at the time that the Active Speakers list should be populated LLVoice stops and starts it self again. ie: If I'm walking around all alone without anyone around with Voice, I can consistently see myself in the Speakers List. As soon as I get near someone that has voice, and the list is going to get updated, LLVoice restarts, that means, the list goes blank and in a while my name comes back to the list.
This doesn't look sound related at all.

Any ideas? Drake?

Royer, if SLVoice does not connect correctly with the network when the sound settings are not correct, it obscures whether its not working because of a network issue or a sound issue (or both).
So maybe this could help to get clearer:
do you have ubuntu, arch and fedora -
1)on the same machine? if no: what soundcards is on which distro?
2)does "sudo iptables -L" (without quotes) in the commandline show the same output on every disto?
3)always using the same connection to the internet? if no: what differs in the way from computer to whatever is connected to the internet? (e.g. one is connected directly with a wire to a adsl-modem the other goes through a router and over wlan)
4)always using the same sl account? if no: does voice work with all sl-accounts on the ubuntu-environment ?

hi all, another tester for voice on linux...
i'm really glad things are getting to work, thanks to developers and early testers!

i managed to have voice working, when it had the alsa device all for him, uncommenting all 3 the "LL_BAD_*" in the secondlife script (i haven't tried yet to talk, but i'm quite confident it will work...).
i installed pulseaudio in place of standard esd, with good improvements for other apps (sl and amarok work fine together at last!..), and wanted to set Vivox to use it. so i made the config file ~/.openalrc with the line: (define devices '(esd))
but then in the configuration window in sl i get only default, oss and null dev, but no esd. obviousy it doesn't work.

this is my experience. what to do?

thanks for attention, bye

OK, so I was able to get voice working in Linux...sort of. I can hear people talking on voice, but I can't talk in voice myself. I've tried a lot of the suggestions made here thus far to get this mic working, and so far nothing's worked.

Info on my setup:
Ubuntu 7.10
Intel HDA sound card (it's a laptop, so I can't go out and install a new one)
ESD as main sound server (tried to replace it with PulseAudio, but got no sound in SL afterward)
SL has all three sound routes enabled -- ESD works fine, OSS works but stutters, ALSA doesn't work at all (presumably it's using ESD)
Headset uses the analog mic input, not USB or anything like that

,openalrc Contents:
(define devices '(alsa native esd null))
(define alsa-out-device "hw:0,0"
(define alsa-in-device "hw:0,0"

esound.conf Contents:
[esd]
# autospawning is not recommended, since it can't really be done
# right. If you want your login session to be using a sound daemon,
# you should start it from the session controller, not some random
# app inside.
auto_spawn=1
spawn_options=-terminate -nobeeps -as 2
spawn_wait_ms=100
# default options are used in spawned and non-spawned mode
default_options=

Currently, I can find no .asoundrc or equivalent in /etc/, and there is none active in my home directory. I assume it's using the defaults.

Any help or other ideas would be appreciated.

I read some place, from the SL team, that for now Vivox will only work with ALSA.
So gotta make alsa work well with ESD. ESD should run before anyother sound thingy, as a service. Works all natural in my up-to-date Ubuntu 7.10.

Do you actually have your Mic working? Can you record sound in linux?
Did you use "alsamixer" to set the correct input device? Yes, the original text alsamixer from the alsa-utils.

Ok, desperate news for me. The whole problem was not with the linux distro.
Just testes Ubuntu 7.10 from scratch, no updates, original 7.10 CD. Works fine in my desktop with snd_intel8x0 (AC97) onboard, but doesnt work in my notebook Dell Latitude 131L with hda_intel sound.

Basically, it's all sound. Works through LANs and firewalls.

But I will never give up. Any clues?
Can someone tell me why the kernel/module sound driver would break LLVoice? Gimme more clues! And why wont LLVoice outputs tell me anything? Yes I have tried the debug tips in the wiki for LOTS of details... but no clues of any problemas!

The mic itself works and is selected as input...I tested it successfully using arecord.

The OpenAL library included with SL may not be compiled for using ESD with Vivox. Remember, Vivox exclusively uses OpenAL.

Also, you may need to specifically use a dmix'ed or dsnoop'ed alsa "virtual device." I'm just letting it directly connect to a Logitech USB headset.

thats really good news drake
if thats the problem with openal then there should be no problem - openal is gpl, isn´t it ? so IF vivox did changes to openal the changes would be gpl, too. then we can just recompile it with esd or whatever useful (pulse ?) enabled.
Better case IF NOT then we can go right now.
I checkout whether vivox's openal just can be replaced with my ubu openal.

btw do you think bluetooth-alsa would do the job of having a second soundcard ? this was cool, i always wanted one of those bt-headsets but couldnt think of any use until now ...

I did and it worked (the new library that is). Replace the one in the secondlife lib dir with the newly compiled libopenal.so.1 (notice the 1). The version has to be the same or it wil break vivox


yes you can. just edit your .openalrc and make sure that it uses the head set.

BTW it is not working on my main system, because of intel HD audio stuff... testing goes on.
On my lt with fc 7 it seems te be working, mic in not testet yet.