locked
Using Bing Translator just for Text to Speech RRS feed

  • Question

  • My WP7 app in the Marketplace uses Bing API to speak output phrases via the SOAP interface at the interface: http://api.microsofttranslator.com/V2/Soap.svc

    In using the Bing Translator (or Microsoft Translator) API in a Windows Phone 7 application, simply to speak phrases (SpeakAsync), as far as I know, I have so far hit no limits using TTS.  What limits will have in the future (after the Sep 2012 deadline)?

    If it what I think it is, none of the tiers are really workable for a mobile app, let alone the free one.

    For instance, at the free tier, I get 2M characters per month.  Seems like a lot until you realize that you need to divide that by roughly 30 to get the number of characters per day, and then by some useful number of characters to get the number of devices.

    Let's say, for a second, I expect a very small 1000 characters per device.  My number of devices I can support on the free tier is about 66.

    Now let's say I get 1000 users, and each of them requires just 1000 characters per day.  I'm near the 32M character/mon level, and that costs me $300/mon, so I'll be able to pay for about 3 months before I'm in the red on my app.

    If I'm wrong, and "plain TTS" has no such limit, I thank you about about one month's worth!

    If not, will Microsoft consider a per transaction limit, instead of characters, for use of SpeakAsync ONLY?

    -THANKS
    -E


    -- me --

    Tuesday, March 20, 2012 4:39 AM

Answers

  • Hello E,

    you see it correctly: the TTS method follows the same volume calculation as translation: the size of the input matters, in characters. It is also correct that a subscription for 2 million characters per month is free. This enables you tp experiment and try the service without having to pay, and at the time of a wider launch you'll get to a volume that moves your app out of the experimental realm.

    We have no current plans to charge for the TTS method differently than for the translate methods.

    Chris Wendt
    Microsoft Translator

    Wednesday, March 21, 2012 9:35 PM

All replies

  • Hello E,

    you see it correctly: the TTS method follows the same volume calculation as translation: the size of the input matters, in characters. It is also correct that a subscription for 2 million characters per month is free. This enables you tp experiment and try the service without having to pay, and at the time of a wider launch you'll get to a volume that moves your app out of the experimental realm.

    We have no current plans to charge for the TTS method differently than for the translate methods.

    Chris Wendt
    Microsoft Translator

    Wednesday, March 21, 2012 9:35 PM
  • Sorry you guys feel that way, but I do appreciate you taking the time to answer.

    With no other viable options, I had selected Bing Translator as my source of TTS, but now I guess I will have to find something else.

    With the volumes as they are on Windows Phone any individual developer these days, the pricing model for this service doesn't lend itself to being used for that sort of work.   I'm sure that a lot of those translator apps will stop working as well.

    Have a good day.

    -e


    -- me --

    Wednesday, March 21, 2012 9:57 PM
  • Hello E,

    can you explain why a per-transaction pricing would be better for your phone app than a per-character charge?

    It still wouldn't be free.

    Chris Wendt
    Microsoft Translator

    Sunday, March 25, 2012 2:47 AM
  • I think so.

    Putting a character limit on it causes apps that want to speak to be more computer-like, and less personable to the user.   The main point of TTS nowadays is to sound natural and not off-putting to the user.

    Compare : 
    "No playlist called My mix."  To
    "Sorry, sir, I didn't find any playlist called My mix in your collection."

    Continuing from that user feature wish, one would have to have less transactions, but if one had a per transaction limit, one could allow the speech to be more verbose without penalty.

    So if I use my example above, and assume something like a 100 transaction limit, I can speak a lot more often without worrying about how much is in each speech snippet.

    This is also more competitive you compare to Nuance' offerings, which are not length-based, but transaction-based (it is unfortunate how much they are lmiited, but they are transaction based).

    Now, let me move to Recognition as well as TTS for a moment.  I have sent communiques to the old "thephone" guy and have attempted to get the attention of JoeB.  I've registered on the TellMe mobile interest site multiple times.  I realize it is a little off topic, but translation just logically needs both recognition and text-to-speech (can't build a universal translator without it ;) )

    It should be possible for developers to use the service to push more people to the Windows Phone platform, and running a TTS service for developer use isn't going to be any worse than the speech recognition and TTS that already exists on Windows Phone.

    I say this because a certain device, no matter how many apps are speech enabled, still has the same "worst case" needs for the services.

    I'm hoping that JoeB and the clan at Windows Phone will put the same speech services in our hands as developers as the phone currently has only for its own interface, but in the meantime, we could provide some help to developers who want to compete with the likes of Siri.

    I don't think it makes sense for 20 apps to each have their own server with different abilities when Microsoft could make sure the platform has one great service behind it.

    -thanks for your consideration
    -e


    -- me --

    Monday, March 26, 2012 12:46 AM
  • Hi E,

    thanks for sharing your views. I appreciate your desire to write compelling Windows Phone apps. The Microsoft Translator API is not restricted to specific platforms, and is designed for broad use in all kinds of applications. It adds to the simplicity of integrating the API that there is a single mechanism to calculate the consumption. A per-character model makes the most sense for the majority of the offered methods, and is in fact most reflective of the actual cost of providing the service. That includes the TTS method.

    I hope and wish that you continue to write interesting Windows Phone apps, and please don't hesitate to let us know at mtcont@microsoft.com whatever we can do to make integration of translation and speech synthesis easier.

    Chris Wendt
    Microsoft Translator

    Monday, March 26, 2012 1:29 AM