WavFrag

Wavefrag is easy to train. Many times a single uttering (pronunciation of a word or phrase) will train WaveFrag. Of course, pronouncing it several times helps. In practical use we found that WaveFrag can compensate for most accents, tired or sleepy voice, room acoustics, microphone differences, sound card differences. There are some limitations to how much WaveFrag can compensate, but if you find that WaveFrag doesn't recognize a particular word, training is as easy as clicking on a button.

To train a new word, click on the 'New Word' button. A small dialog shows up expecting an input string. Enter the name of this word. The name doesn't have to match the pronounced word, but it helps in identifying it later. After clicking okay, the word is added to the 'Trained Strings' list. With WaveFrag listening, pronounce the word. The wave the window (on top) will show the wave form of the pronounced word. Click on the 'Train this wav' button, and your newly trained word is remembered. Pronouncing it again, and clicking on the 'Train' button will train the word more, reinforcing your speech pattern. training at two or three times will train the word into perfection.

To monitor if you are satisfied with the current uttering, you can always play back the current wave. A word of caution, the sound coming from your loudspeaker will be recorded again if you are on a hot microphone. In that case, instruct WaveFrag to stop listening, so you can freely monitor or review the current uttering.

To train an existing word, select the desired word in the 'Trained Strings:' list box, pronounce the word, and click on the 'Train this wav as' button. (On the screen shot above, we clicked on 'Activate Next', and the training button shows "Train this wav as 'Activate Next' ". To review an existing word, select the desired word in the 'Trained Strings:' list box, and click on the desired wav in the 'Waves for this string' drop down box. the waveform should show in the top wave editing window, one can play, edit, or crop the selected waveform. (Don't forget to save it) To delete an existing wave, click on the 'Trained String' you desire, and in the 'Waves for this string' list box, click on the wave you wish to delete. Once a wave is selected, click on 'Delete Entry' button. A confirmation dialog will ask yes or no. Answering yes will delete the wave. To delete an existing (trained) word completely, select the desired string in the 'Trained Strings' listbox, then click on the 'Delete Word' button. A confirmation dialog will ask yes or no. Answering yes will delete the string, and all of the wave files (and training) associated with it. Deleting the word cannot be undone, however it is easy to retrain it.

Once a word is trained to your liking, you can add it as an action. An action can execute a program, show an image, switch between windows, send keystrokes to windows, or even shut down your system. In short, WaveFrag can be configured to do almost anything with a computer.

Clicking on the 'Add' button will open a small dialog to ask for the name of the action. WaveFrag will pre fill the action name with the name of the last uttering, but one can name the action to any name desirable. Next, clicking on the configure button will call up the action configuration dialog. this dialogue is described in more detail in the commands section of this manual.

One can edit any pre-existing action. Selecting an action on the drop down box, and clicking on the 'Configure' button will allow one to change that action. There are pre-made actions to close windows, to switch between windows, to show images. None of those actions are hardcoded, one can change any and all of them to a custom command driven by a custom uttering. In fact, WaveFrag can be trained in a foreign language as WaveFrag has no knowledge about lexical or phonectic properties.

WaveFrag is very versatile when it comes to commanding your computer by voice. One can train any word, any uttering, in fact any noise to be recognized by WaveFrag, and acted upon.