FOOD FOR THOUGHT Replacing GoogleVoice voicemail TTS...

[bobkoure]

... turns out to be pretty easy
- get a free account on Nexiwave
- set an extension to send voice mail to your nexiwave address ([email protected])
- and that's it (voicemail mp3 files plus transcript get sent to the address you registered with)

The hard bit was getting postfix to relay through gmail, as gmail requires TLS.

Nexiwave has an API. Ran across it while building a GoogleScript that uses the Google TTS API. But AFAICT the Google API is undocumented, meaning it could disappear tomorrow - and the Nexiwave setup was dead easy.
Note to Ward: please consider pre-installing and configuring postfix on future incrediblepi builds (let users just add gmail username and password). I'm guessing there'll be a bunch of folks looking to replicate what GVoice does on RasPi and BeagleBone boxes in a couple of months.

 

[Eric Walker]

There is now an Asterisk Integration Guide from Nexiwave. It's slightly different from Bob's email proxy based method (which is also great): it posts the audio to Nexiwave via HTTP API, receive the raw transcript back and then send email as usual. There is no need to mess around with postfix or gmail.

See here: http://nexiwave.com/index.php/146-nexiwave-voicemail-to-text-asterisk-integration-guide

Ward, this might be worth to put into the pre-install.

 

[wardmundy]

Great idea! Thanks.

 

[Brian Simmons]

Nexiwave only supports wave file for now. I guess that means you have to decide if you want to get TTS, or a MP3 playable on a mobile device.

 

[wardmundy]

Stay tuned

 

[Eric Walker]

it seems Ward has worked with nexiwave... mp3 is now supported.

 

[bobkoure]

Ward - I owe you at least one beer. Coming to Boston anytime soon?

 

[bobkoure]

Has anybody gotten this up on incrediblepi?

Packages are installed.
nexiwave_v2t_mail.sh coped and chmod'ed

It's already setup to send mp3s (mailcmd=/usr/sbin/sendmailmp3), which I think might be one of Ward's customizations.

BTW, sendmailmp3 is set via FreePBX admin / settings / voicemail / settings / mailcmd - which makes me think that this file will be modified if any changes are made on that page)

I've tried putting
mailcmd=/etc/asterisk/nexiwave_v2t_mail.sh <nexiwave username> <nexiwave password>
before mailcmd=/usr/sbin/sendmailmp3, after it, and with it commented out.

No voicemail, with or without transcript, is coming through.

Any ideas?

Thanks!

 

[Trimline2]

I think you need to edit voicemail.conf. I haven't installed it as I don't have a need.

See the steps here: http://nexiwave.com/index.php/146-nexiwave-voicemail-to-text-asterisk-integration-guide

 

[bobkoure]

I did that. Three different ways. (add the mailcmd statement before the existing one, after the existing one and replace the existing one).
None of them seemed to work - so I thought I'd ask here.

Not a biggie, works fine when I relay email through them rather than doing it this way. I was just hoping for a bit more control over the email subject field when it gets sent to my mailbox. I can probably switch it to what I want with gscript; this just looked a bit less of a kludge.

 

[visionlogic]

I did that. Three different ways.

As far as I can tell the "mailcmd=" can only contain one object. You should only need to use "mailcmd=/etc/asterisk/nexiwave_v2t_mail.sh <nexiwave username> <nexiwave password>" to do it all. The conversion of wav to mp3 instructions contained in sendmailmp3 are included within the nexiwave_v2t_mail.sh script. Ward supplied the info to Nexiwave.

 

[Eric Walker]

bobkoure... does your setup work with the stock sendmailmp3? as far as I can tell, the nexiwave script is essentially the same as my original stock piaf one.

how about the "format" in the general section in voicemail.conf? I have it as wav.
i also assume you have all apps installed...

 

[bobkoure]

My system works fine with the stock sendmailmp3, once I'd installed and setup postfix. Major challenge was getting that working with Google's SMTP servers as they want TLS - probably easy for someone who speaks Linux "native". I had a good working knowledge of unix in the 80's, but that's faded, and the details are mostly different.

 

[bobkoure]

BTW, (sorry for not just editing on a ps but edit-message not working for me) if you are following the nexiwave setup, the items for installation via yum are already on your system (needed for sendmailmp3). But yes, I ran apt-get (raspbian doesn't have yum) just to make sure.

 

[Robert-BCC]

The pricing advertised on the Nexiwave front page boasts that the prepaid pricing plan are "~$0.18/min". I signed up, and here are the three prepaid pricing plans:

• $20 for 1 hour (33 cents per minute)
• $250 for 20 hours (21 cents per minute)
• $1,080 for 100 hours (18 cents per minute)

Note that they charge in increments of one second, which is cool. If you buy the mid-tier plan and if 10 users each get two minutes of voice mail a day, then you'd be spending:

• Two minutes = $0.42 /day
• 10 users = $4.20 / day
• 4 weeks = 20 days
• So that's $84 a month for voicemail transcription for a small office.

Given that Vitelity and RentPBX are hosting servers for $20 - $30 per month, and call minutes are around 2 cpm or less, this would be a pretty difficult service for me to justify.

 

[Brian Simmons]

Well, to be fair the pricing on the Nexiwave front page does say, "**: Actual price varies based on the purchase amount. See Purchase Page for exact pricing (requires login)."

 

[Robert-BCC]

Not revealing your true pricing up front to people you want to be your customers ... maybe it's a great idea. Heck, it works for car dealers and we all know how much we enjoy working with them.

Anyway, my beef isn't about the disguised pricing. It's about the expected monthly cost of the service, which is greater than the combined cost of paying to have someone run an entire PBX and the cost of thousands of call minutes.

 

[krzykat]

Not revealing your true pricing up front to people you want to be your customers ... maybe it's a great idea. Heck, it works for car dealers and we all know how much we enjoy working with them.

Anyway, my beef isn't about the disguised pricing. It's about the expected monthly cost of the service, which is greater than the combined cost of paying to have someone run an entire PBX and the cost of thousands of call minutes.

Just out of curiosity because I'm working on a very similar project that I've been working on a while and getting ready to release.
What would be a "fair" price that you think would get what you need?

 

[Robert-BCC]

Ask anyone who manages IT budgets and they'll tell you they're looking for predictable service costs. So I'd support a model with a fixed monthly price on a per-IP basis. That is, one server has the right to request as much transcription service as it needs for a given month. As for what's a fair price, that's harder. Given Nexiwave's stated price lowest per minute cost, one IP constantly (serially) requesting service could consume over $750 / month of transcription.

But me? I could see justifying a $10 - $15 / month charge for a single IP.

 

[Eric Walker]

I'll say fair price is relative. Up till early this year, we had to listen to voicemails and type into our ticket system... more than a fulltime job during busy campaigns. We are pretty happy with nexiwave's $250 option... the voicemails now go into the ticket system directly as text... certainly worth the money.