VoiceDial foriPhone is speaker-dependent and requires a short period of training. It may then successfully capture your speech with a very high accuracy. It can achieve 96% accuracy if operated under optimal conditions.

It needs to learn your voice and can only recognize previously recorded items. It does not have a predefined vocabulary, so you need to record the words and names it recognizes.

You may start VoiceDial by tapping the icon on the main iPhone menu. For fast dialing you may drag this icon to the bottom row.

Firt, try recording a few names from your addressbook. You have 2 seconds recording time for each name (called 'voice tag'). You may record only the first or last name of a person, if you wish. If you record both, skip the pause between the first and last name. For example, say "JohnSmith", not "John...Smit" for better results.

Press Record New \ Contacts \  Name \ Number to start recording

For example: Record New \ Contacts \ John \ (1)23456789

When you go to Recordings, you see all the contacts for which you have stored speech samples. You may also add the three confirmation commands (see Preferences section below)

If you start the application, the software will say: "Speak Now" and starts capturing your voice for 2 seconds. It will then try to match the sound sample to the ones recorded earlier.

When it finds a match, it will playback the sample that was stored for that action, then dial the number.

For the best recognition results follow these guidelines:

1. Record clean samples. After each recording, the software plays back the captured sample. If you hear crackles, pauses, distortions, background noise or interruptions: don't save the sample, but cancel it. Then repeat the recording until you have a clean sample.

2. Record multiple voice tags for each action. You may store up to 3 voice tags per action. More voice tags mean a better chance of a positive recognition. Go to Commands and choose an item from the list. Then select Add Voice Tag to record additional tags.

3. When recording multiple voice tags: apply slight variations in the recordings. Use a different pitch for the second voice tag and a different tempo for the third.

4. Less is more. If you store too many contacts, the recognition accuracy will drop, because you will have more people with similar sounding names, making it too difficult for the software to distinguish between them. To get 95% recognition rate, store the 10 phone numbers you dial most. You may store more, but the recognition rate will drop considerably.

5. When recording names: record the first name and the last name without a pause . Our speech recognition algoritm doesn't like pauses between words.

6. When recording names: avoid only first names with single vowels. Short names, like John and Josh, or Beth and Ben are phonetically very similar and the software may mix them up. Always say the last name as well in such cases.

7. In the recognition phase: make sure you're in a clean noise environment with as little as possible background noise. For example, while driving: turn off the car radio and close windows.

8. Make sure your hand is not covering the microphone during recording or recognition. The iPhone microphone is located at the bottom right of the device. When covered, the sound sample will be muffled.

9. If you have frequent mismatches of a name, reboot your iPhone to make sure other processes on the phone are not interfering or taking away processor time.

10. If you still have mismatches after a reboot, add three new voice tags for that name. The old ones will automatically be overwritten.

Confirmations - when enabled, you'll be asked to confirm every action before it's performed. When turned off, it speeds up dialing by approx. 5 seconds. To use this feature, you must record the 3 basic confirmations commands: Accept , Repeat & Cancel . You need to do this before you record other samples. Go to the Record New , then select Confirmations, then record all three commands.

When enabled, you'll be asked for confirmation: "Accept, Repeat or Cancel". Accept will perform the action. For example, it will dial the default number of the person's name whom you have spoken. Repeat enables you to speak again after a false recognition. Cancel will stop the recognition process.