vicidial.org

We have a new cluster implemented for our collections group. We are now setting up Cepstral TTS 5 channel licenses, 2 voices and started making the recording prompts for it. Here is my question. We have 3 telephony servers clustered 'ast1' 'ast2' and 'ast3'. Ast3 is going to be running the virtual campaign with TTS and this is where the Cepstral software and licenses are installed. Do the recording prompts for it go on the sounds web server for the cluster or on Ast3? In other words will the TTS recognize the cluster sounds location or do they need to be on Ast3 for TTS to play them?

Update: We implemented and tested. Silence. We have run this in a single server environment before but when we implemented this install into our cluster, I cannot get it to play the TTS from survey. Just Silence. I tried copying the TTS generated audio files to both the web sounds server and 'ast3' that has the TTS components installed and licensed. It also did not 'say' the --A--first_name--B-- --A--last_name--B-- fields.
Has anyone used Cpestral TTS on a cluster?

CLI From test call.

[Jun 14 16:27:03] -- Executing [912107222479@default:1] AGI("Local/912107222479@default-23bd,2", "agi://127.0.0.1:4577/call_log") in new stack
[Jun 14 16:27:03] -- AGI Script agi://127.0.0.1:4577/call_log completed, returning 0
[Jun 14 16:27:03] -- Executing [912107222479@default:2] Dial("Local/912107222479@default-23bd,2", "SIP/voipinnooutast3/12107222479||tToR") in new stack
[Jun 14 16:27:03] -- Called voipinnooutast3/12107222479
[Jun 14 16:27:06] == Parsing '/etc/asterisk/manager.conf': [Jun 14 16:27:06] Found
[Jun 14 16:27:06] == Manager 'sendcron' logged on from 192.168.25.223
[Jun 14 16:27:06] == Manager 'sendcron' logged off from 192.168.25.223
[Jun 14 16:27:13] -- SIP/voipinnooutast3-00000004 is making progress passing it to Local/912107222479@default-23bd,2
[Jun 14 16:27:15] -- SIP/voipinnooutast3-00000004 answered Local/912107222479@default-23bd,2
[Jun 14 16:27:15] > Channel Local/912107222479@default-23bd,1 was answered.
[Jun 14 16:27:15] -- Executing [8374@default:1] Playback("Local/912107222479@default-23bd,1", "sip-silence") in new stack
[Jun 14 16:27:15] -- <Local/912107222479@default-23bd,1> Playing 'sip-silence' (language 'en')
[Jun 14 16:27:15] WARNING[22940]: file.c:1297 waitstream_core: Unexpected control subclass '-1'
[Jun 14 16:27:15] -- Executing [h@default:1] DeadAGI("Local/912107222479@default-23bd,2", "agi://127.0.0.1:4577/call_log--HVcauses--PRI-----NODEBUG-----16-----ANSWER-----12-----0") in new stack
[Jun 14 16:27:15] -- Executing [8374@default:2] AGI("SIP/voipinnooutast3-00000004", "agi://127.0.0.1:4577/call_log") in new stack
[Jun 14 16:27:15] -- AGI Script agi://127.0.0.1:4577/call_log completed, returning 0
[Jun 14 16:27:15] -- Executing [8374@default:3] AGI("SIP/voipinnooutast3-00000004", "agi-VDAD_ALL_outbound.agi|SURVEYCAMPCEP-----LB") in new stack
[Jun 14 16:27:15] -- Launched AGI Script /var/lib/asterisk/agi-bin/agi-VDAD_ALL_outbound.agi
[Jun 14 16:27:15] -- Playing 'sip-silence' (escape_digits=) (sample_offset 0)
[Jun 14 16:27:15] == Spawn extension (default, 8374, 3) exited non-zero on 'SIP/voipinnooutast3-00000004'
[Jun 14 16:27:15] -- Executing [h@default:1] DeadAGI("SIP/voipinnooutast3-00000004", "agi://127.0.0.1:4577/call_log--HVcauses--PRI-----NODEBUG-----0---------------") in new stack
[Jun 14 16:27:15] -- AGI Script agi://127.0.0.1:4577/call_log--HVcauses ... ---------- completed, returning 0
[Jun 14 16:27:16] -- AGI Script agi://127.0.0.1:4577/call_log--HVcauses ... --12-----0 completed, returning 0
[Jun 14 16:27:16] == Spawn extension (default, 912107222479, 2) exited non-zero on 'Local/912107222479@default-23bd,2'

I see where it is playing 'sip-silence'

First off verify that the audio files are mono 16bit 8kHz PCM wav files. If they are not Cepstral can bomb out.

Second off please go into the Server record for Ast3 and make sure that the AGI Output is set to BOTH. Then go into the /var/lib/asterisk/sounds/tts directory and delete everything in there. Then place a call. After that go into the /var/log/astguiclient directory and run this command:

Code: Select all

less -r screenlog.0

Scroll down till you find your test call and post everything from the beginning to the end here. Also post the full TTS prompt you are using.

We got it setup correctly!
FYI, you can have one telephony server running Cepstral TTS in a cluster of telephony servers. All Cepstral actions and functions are performed on that one server. If you want to use Cepstral output to file, 'make sound files with say Allison-8kHz' do it on the server with Cepstral installed, then transfer the sound file to the server and location that is running your sounds web directory. Then those can be used as prompts in any audio chooser field.

FYI, you can have one telephony server running Cepstral TTS in a cluster of telephony servers.

If all you are just using Cepstral TTS to generate static prompts then yes, otherwise no. If you are going to be using TTS to do things like say the customers name in a Survey campaign, then you need to have Cepstral install on all of the dialing servers.

Let me explain how we are doing it. We have Cepstral TTS in production on only one of the clustered telephony servers. All other telephony servers have Campaign Trunk Setting of 0, Maximum_Limit to for the survey TTS campaign to only dial out on the telephony server with the Cepstral licenses and software installed. This way we are able to reduce our Cepstral license expense.
This is in production now and working beautifully.

Yes that would work fine, but that is because the TTS calls are only going out that one server.

That is correct. Which is what we want. We are also going to test Sangoma CPD for this server's VOIP endpoint. This way we will not have a huge license investment for this either and with the custom bash scripts we have developed for the call results, 'These email us bad numbers' we can clean up ans skip our debt accounts.