SpinVox: Now All I Want Is Noodles

For the longest time, the only United Kingdom company that I wished was in the United States was Wagamama because I love noodles. (Alas, it’s finally coming to Boston, but I go to London more often than Boston.)

My list of UK-envy companies has doubled because SpinVox gave me an account. This company’s service converts voicemail to text and then sends the text to your phone or email account. Now I don’t have to listen to voicemails at 650-555-1212. I just read the email at guy@garage.com or get a text message.

At first I was skeptical about the accuracy of the conversion, but I’ve been very impressed. Here’s an example. I left myself a voicemail of the previous paragraph, and this is the unedited text that was sent to my email address:

“My list of UK handy (?) companies has doubled, because SpinVox gave me an account. This company service converts voicemail to text & then sends the text to your phone or email account. Now I don’t have to listen to voicemails at 6505551212. I just read the email at guy@garage.com or get a text message.”

There are roughly 250 million cell phones in the US, and they generate roughly 90,000,000,000 voicemails (250 million x 30 voicemails/month x 12 months). And like any good entrepreneur knows, you only need to get 1% of a market to be successful. :-)

The bad news about SpinVox is that it’s not yet available in the United States. As you can imagine, the company is conducting trials and discussions with carriers all over the world, so if you know any executives at carriers, you should urge them to add the service.

I am a simple man: I just want to eat noodles at a Wagamama in Palo Alto while receiving the text of voicemails. I’m half way there, and I may never answer my phone again.

---

Comments

[dendar]

SpinVox seems to use a combination of speech recognition and human QA from what i can see from all the articles and interviews. They also seem to have dozens of patents in and around this. Its a very difficult thing to do but they seem to have cracked it on the basis of all the carrier deals they have managed to do.

[Ex-Spinvox Transcription Operator]

Spinvox does not have an automated transcription system. All messages are transcribed by human operators located at various locations around the world. How is it possible that speech software for your personal computer takes ages to recognise your own voice yet a company has this uber-smart system that can perfectly transcribe strangers voices. All transcription is carried out using their own transcription software called Tenzing (used to be called Fruitbat). This allows the operator to select the language, the message queue and to rewind and replay the message until it is transcribed. The transcribed message is then sent via an in-built smtp engine to Spinvox servers in the UK (Maidenhead) where it is transmitted to the recipient. Simple really.

Spinvox will do anything to keep this secret as their 'uber-cool transcription system' makes them a highly valuable technology company. Messages transcribed by humans is nothing special and they are devalued.

You will notice that some common English names are transcribed incorrectly where as Indian names are always perfect...

[Read The Patent]

As the patent states, SpinVox is human-based and not technology-based.

Hence, not the most scalable of businesses.

No wonder they've had problems trying to sell it...

http://www.freepatentsonline.com/20060223502.html

[Shahed Moolji]

Look at their patent.

They clearly state that humans transcribe !!!

http://v3.espacenet.com/textdes?DB=EPODOC&IDX=GB2420943&F=0&QPN=GB2420943

[Jack Cage]

Hi. GREAT blog.

Where's the money to be made in v-to-text? 1. What do the carriers make? 2. What do these companies make? 3. Can the text be captured and targeted for marketing purposes...like Google does today? That is, capture tel#s, email address, URL (Website incl. MySpace, Facebook, etc.).

thanks, jack

[n/a]

I've been told by an insider that the Spinvox human transcription centres are in Ireland and Barbados. The insider also told me that the're frantically working on automated conversion using 'naive' venture capital cash but it's a monumental task (impossible ?)

Was also told that their system is Linux based running the free Asterisk voip system - the've: a) converted the Asterisk US voiced menu .wavs to uk ones b) moded Asterisk so that voicemail emails are auto sent to Ireland when received c) humans transcribe then send back email replies d) Spinvox send out received transcribed text via a uk sms carrier

[180srch]

I'm surprised my earlier post was deleted. Was it due to the fact that I mentioned our service which uses text messaging, rather than the aforementioned voice system (jott, simulscribe, dictabrain) ???

[n/a]

I've read about a service jott.com in the US.

[Keith Mitchell]

I've been using spinvox for over a year now, and it is utterly indispensible. I have impaired hearing, and it has often been the difference between understanding important voicemails and not. It is also vastly quicker to skim through your messages as text/email than have to listen to them all. Since moving to the US, it has been a major dissappointment they don't work with any of the US cellular operators, to the extent I have an elaborate setup with VoIP virtual numbers to use it from here. There is a major opportunity for both cellular and VoIP operators to work with companies like spinvox.

[Frank]

There's a similar service in Beta called Jott which does a similar thing. Niklas Zenstrom of Skype is an investor. Give it a whirl.

[PAUL]

Passionmantra. "Why read rather than listen to voicemails" you ask. I could give you my list but try this for starters..I save 5 hours /month not listening to Voicemails.Nor do I have to write down phone numbers anymore Thats a 2.5% effeciency gain . I'd take that even without all the other benefits.

[PAUL]

Now I know Spinvox has what it takes. I checked out Wagamama last night in London.You were right,noodles of delight. Informing the manager of the blog and how he was double billed with Spinvox he left me a voicemail. Here is the Spinvox text ,it's exactly what he said....... you can't fool Spinvox with Wagamama it seems.

---------------
Mr. Paul, thank you for coming to Wagamama(?) for your noodles(?). I'm pleased you've enjoyed yourself.
- Powered by SpinVox.
---------------

Message received at 06/12/2006 21:57:42

[passionmantra]

nice article, something that struck me so hard. still thinking...why would someone want to read vs listen to voice mails?

[Shefaly Yogendra]

No matter who is transcribing, they can at least spell and I am relieved to see decent spelling is (mostly) alive and kicking. Someone has to innovate swiftly before the text-messaging generation gets to this and the message reaches your mobile reading as follows:

"My lst of UK handy (?) cos. hs dbld, cuz SpnVx gv me an a/c. Ths co srvc cnvrts vmail to txt n thn snds da txt 2 ur phn or email a/c. Nw i dnt hv to lsn to vmails @ 6505551212. I jst read email @ guy@garage.com or gt a txt msg."

[Steve]

Guy,

A similar "carbon based" (aka Human) service has been here in good 'ol New Zealand for close to 4 years. Aangel messaging goes one step further - appointments, tasks and contact details spoken down a short code "while you are on the highway" and sent back via text (vCal) and email to your phone, calendar or organiser.

It is the hottest service here being used by the #1 mobile carrier here.

No more forgetful moments. And when was the last time you used that calendar function on that Motorola Razr you use?

These guys are on the look out now for international partners...

www.aangel.com

tell them simmsy sent you...

[RJB]

Wagamama!!! How I miss it!!

Have you ever tried Satsuma?

[PAUL]

I'm an avid SpinVox user and know the service is life changing it makes laborious, and frankly awkward, voicemail a breeze - one glance and I'm instantly in-touch and can respond. In fact, it's more like being spoken an SMS or indeed an email which is actually what I want and someone calling me wants - get me the message. It's exactly what they do - don't leave a voicemail and send me a text, so SpinVox is bang-on. And they've really cut it with product/service design - it's so slick and well thought out, you can't help but be in awe of it

I happen to know the industry (ex Motorola VP and General Manager of European Cellular) and met with 'insiders' from both carriers and SpinVox. I know they're not using large pools of humans, but developed a very smart system that only occasionally uses the odd human for difficult parts of messages. But it then learns. That's the point of humans they enable machines to be smarter and business’s to be scaleable. They've invested a ton in their state-of-the art conversion system and use speech recognition in very novel ways. Carriers who are deploying SpinVox have done their due diligence and were impressed, so I know it's a WOW for them. And these are big carriers both in Europe and state-side, so I'm expecting them to storm this market.

I see mention of copy-cats who are probably using pure human conversion, like Simulscribe who openly admit it. Don't blame them. It's such a great idea, who wouldn't want to be in this market. Their site is eerily word-for-word copy of SpinVox's...

If I find the Achilles heal I will let you know but I haven’t found it yet.

[Eric]

Heh. Came here to add a note about simulscribe.com but I see others have beat me to it. Couldn't agree more about Wagamama, btw... best, E

[James M Woods]

Hi Guy,

Thanks for pointing this out. There is a service exactly similar to this in North America called DictaBrain. Launched in Alpha in September at a local Toronto DemoCamp. By all accounts it appears that our services operate in a very similar fashion. There is a demo on YouTube of the presentation (tag DictaBrain) and you can sign up for an account from www.dictabrain.com

Would love to get your feedback.

James

[charlotte batt]

Love the blog!

I know these guys quite well at SpinVox and they definately dont have people in India or South Africa. My understanding is that the conversion is mainly automated with human QA/QC to make sure the quality is always good. I think companies like simulscribe are purely human operators as far as I know.

There are quite a lot of people using the service in the UK now and all the feedback I have read seems to be really positive. I have just put it on my phone and its really cool. love it.

[John Wilson]

Sorry Guy - in my previous post I made an unforgivable error. I meant to say a South African call centre, not Indian. Also, I am assured that Christina, the owner of Spinvox, has "a serious tech roadmap", which is clearly a relief, particularly to the South Africans typing most of the drivel heard in voicemails.

[Leslie McCue]

Enter Simulscribe (www.simulscribe.com) same as SpinVox but works in the US...once you use, you don't know how you lived without...

[Chris Knott]

A decade and a half ago I lived in the West End of London, just a short walk from the original Wagamama.

It's the only food from London that I missed, but then I left just before London turned into Foodie heaven.

Now I live in the North End of Boston (spoiled for restaurants) and I'm elated that I'll be just a short walk from the States' first Wagamama. Mmmmm.

As for SpinVox - I'd easily pay $10/month on my cell bill for the service, but I'd love it if my work phone could somehow be hooked up too...

[Bob King Neverland III]

"I’m half way there, and I may never answer my phone again."

Oh man I couldn't agree more.

[John Wilson]

John Dodds is absolutely right about it being a human translation service. I understand that the transcription is being done out in India. I wrote about this back in July under the heading "More Spin than Vox" (http://greatapps.blogspot.com/2006/07/more-spin-than-vox.html). They were/are looking for about £30m funding to build greater automation and recently hired a CTO to direct this work.

Doesn't negate its' usefulness, merely removes the "Wow" factor.

***********

John,

What source told you this? You're absolutely certain this is true?

Thanks,

Guy

[John Dodds]

While on the infamous Scoble pub crawl last week, I was told that the transcription is done by humans not technology and, if that is true, I imagine that potential privacy issues might concern you.

[Ian Eiloart]

Hey, if you're going to use the "555" area code to avoid publicising real phone numbers, then use the "example.com" email domain which is reserved for the same purpose.

see rfc2606, section 3

[Jordan Mitchell]

Mmmmm ... Wagamama. I'm with you, having recently moved back to Seattle from London. I ate Wagamama at least once per week.

[Anand]

Well spinvox was reviewed by techcrunch way back in May 2006 and a similar service but that converts a voicemail to mp3 is gotvoice.com..
both reviewed by techcrunch : http://tinyurl.com/yh373p

*************

Anand,

SpinVox was mentioned by TechCrunch. Michael has not yet used it. I am using it.

Thanks,

Guy

[jeff barson]

I use copytalk in almost exactly the opposite way, to call in to post to my blog: http://www.nimbleit.squarespace.com/the-blog/2006/7/3/post-directly-to-your-blog-using-the-ultimate-speech-recognition.html

I think they use low cost labor in India to transcribe.

[Chris]

How can I concentrate on SpinVox when the same post mentions Wagamama??? Been to one in London - great place! Long tables, orders being radio'd in, great atmosphere. A lot of interesting things going for it - think there is probably a case study in there somewhere re. branding...

[Joe Suh]

I recently blogged about the number of filters we're adding to our phones - voicemails, caller id's, and now voice-to-text! The goal seems to be to NOT answer our phones anymore...

Long after network effects kick in and everyone is on the network (whether it be phones, fax machines, or myspace), we build up immunities and filters to these services. We want to be reachable but not bothered, and thus begin an arms race of filtering out the spam and noise.

ps. Guy, for your noodles, I'd recommend Ryowa in Mountain View :)

[Kempton]

I took a quick look at a few reviews and it is an interesting service, quite useful in many cases. (I thought someone actually transcript the voicemail but it seems like it is automatically converted by machine.)

Mind you, I can see some special messages where I want to hear the voice and the tone of how a message is being left. And in those case, it will be good to refer back to the original voicemail.

Anyway, this seems like a neat service assuming the accuracy rate is as advertised. Speaker independent voice recognition in noisy environment used to be a big challenge, so may be people in SpinVox has it sorted out. Neat.

[Michael Jensen]

I'd be interested in how SpinVox handles names of people, which would be probably the most important part of the message. It sounds like the phone number and email address were okay, but what if it was something like a university email account, axybc9@someschool.edu.

Love your blog!