Technology

  • What is the typical use case of CeedVocal SDK?
    CeedVocal SDK enables you to add speech recognition to your iOS application. Thanks to this technology, your app is going to recognize words from a predefined list while someone is speaking. As soon as a word is recognized, the app will play an animation, open a webpage, execute a command, etc.
    However, CeedVocal SDK is neither for speech synthesis (text-to-speech) nor for dictating someone's speech.

  • Which languages are available?
    6 languages are currently available: English, French, German, Dutch, Spanish and Italian.

  • Do you plan to release additional languages?
    No, we don't at this time. Each language requires a specific acoustic model and phonetizer as it involves specific sounds (called phonemes) and/or letters. We need recordings of several dozens native speakers to build an accurate acoustic model; there should be men and women, younger and older people, soft and deep voices...

  • Is CeedVocal SDK only for iOS devices?
    Yes, it is. Our development kit was optimized for the iPhone and iPad. There's no version of CeedVocal SDK compatible with OS X or other mobile platforms (Android, Windows Phone...).

  • What is the sampling rate used by CeedVocal SDK?
    The sampling rate is 16 kHz. That's why CeedVocal SDK doesn't support audio acquired using 8 kHz mics (like Bluetooth mics for instance).

  • What are the memory requirements?
    • The footprint is around 5-10 MB of RAM for 200 words.
    • Data files are around 1 MB per language, and the linked static library size is around 5 MB per architecture (armv7, i386 available).

  • What can I do if there are more than 5,000 words to be recognized?
    Beyond the 5,000 words limit, launch time and processing time can be too long on a 3GS iPhone. This can be faster with newer iPhone models.
    What you could do if your list is longer than 5,000 words, is to implement steps in the speech recognition process. And each step has its own list of words to be recognized. Think of word categories, followed by actual words.

  • Does CeedVocal SDK support keyword spotting?
    Yes, it does. Keyword spotting is a technique for finding one or several words from a predefined list in a sentence. CeedVocal SDK also includes a garbage model which ignores all other words (i.e. words which are not on the predefined list).

    Imagine a travel guide app. You list all places worth visiting as keywords. For San Francisco, the list would be "Golden Gate Bridge", "Alcatraz", "Lombard Street" and so on. When the user says "I want to go to Alcatraz", the speech recognition technology detects the word "Alcatraz" in the sentence, and ignores all the other ones. As a result, it will show the information about Alcatraz.

  • What is the difference between isolated words recognition and keyword spotting?
    • Isolated words recognition: The recognizer will select the word in the list that is the closest to the one pronounced. It also gives you a score that can be used through a heuristic to determine whether that word was actually spoken or not.
    • Keywords spotting: this mode will recognize a word of the list even if it is within a larger sentence. This mode is less accurate than recognizing isolated words, and takes more time to process. But it's more flexible.

  • Is CeedVocal SDK able to determine who's talking from a set of existing recordings?
    No, it isn't. Locutor identification is not available at this time.

  • Does CeedVocal SDK support speech recognition from an input WAV file?
    No, it doesn't. WAV input is not supported. CeedVocal SDK uses standard iPhone audio queue as input (16-bit, 16 KHz, mono).

  • Can CeedVocal SDK recognize the right sentence from pre-recorded audio files?
    No, it can't. This is another approach, called Dynamic Time Warping (DTW), which is specific to the locutor who recorded the speech. That is not available in CeedVocal SDK.

Trial Version

  • I didn't find any download link for CeedVocal SDK.
    CeedVocal SDK isn't a standard product but a custom-made kit featuring the languages each developer is actually interested in. Therefore, you can't download it as is from our website. If you want to try the speech recognition SDK, write to
    .

  • How long is the trial version?
    The trial version is usually 3 months. However, we know that developing an app often takes more time. So, if you need to extend the trial version after the first 3 months, contact us and we'll issue an additional 3-month trial version.

  • I would like to evaluate the SDK. What will I receive and how?
    After signing the NDA, we'll give you access to a link for downloading a package including CeedVocal SDK (both for iOS devices and simulator), the acoustic model for the selected language, a sample app and documentation.
    You'll also get a time-limited CeedVocal license by email.

  • What's the difference between the trial version and the licensed version?
    First of all, the trial version is time-limited (usually 3 months). However, if the app is not ready after 3 months, you can of course ask for extending the trial period.
    Secondly, once CeedVocal SDK is integrated into your app, you'll see a message to remind you when the license expires.
    Finally, you can't use neither the Ad Hoc distribution, nor the App Store distribution with the trial version.

Commercial License

  • Does the licensing scheme include royalties?
    No, it doesn't. You purchase a commercial license for your iOS app as a fixed fee. You don't have to pay neither royalties, nor commissions, nor annual subscription.

  • I'd like to purchase a commercial license. How can I actually pay?
    You can do it 2 ways:
    • PayPal: it's the fastest way. So if you need to get the SDK within 24 hours, you should use PayPal. Processing costs are around 4% of the transferred amount.
    • International Wire Transfer: it usually takes a few days to transfer the money, and processing costs may vary (it's cheaper if you are located in the European Union than in the USA for instance. And sometimes it's even cheaper than PayPal).

  • I am a contractor/freelancer, and I want to purchase a commercial license for a client. How can I achieve this?
    It happens very often. The CeedVocal SDK License is provided to the final customer who actually submits the application to the App Store. When our direct customer is a consulting company or individual, a license has to be bought by each final customer. We simply mention both your name and the final customer's name in the license agreement.

  • What is the Bundle ID?
    It is a namespace for your applications. For example: com.mycompanyname.MyApplicationName. It uses a reverse domain name sort of format with the idea that only one person can own it at a time.
    We use the Bundle ID to link CeedVocal SDK to your app.

  • Should I send my app for approval to you before submitting it to Apple?
    No, you shouldn't. We don't ask to review your source code either. But it would be great if you could contact us when your app is available on the App Store. So, we could give it a try, add a link to it from our website, or tweet about it...

  • I bought an SDK License for developing a single application but now I would like to develop additional ones. Is it possible to upgrade?
    Yes, it is. Tell us what you would like to do, and we will modify the first agreement you signed accordingly.
    From the financial point of view, you will only pay the price difference between both options. It means that the price will be exactly the same as if you had chosen from the start to develop an unlimited number of iOS applications.


En | Fr | De | Nl | Es | It