Improving performance

Nuance Vocalizer offers several direct ways to intervene in the pronunciation of text with the goal to improve overall performance. They include allowing you to:

• Rewrite the orthography: Without tuning, some TTS audio will not have optimal sound. Experiment with different ways to phrase the same message (for example, using synonyms or changing word order), and you can often obtain a better result.

  You can do this by re-writing static input text, but even dynamically generated text can be rewritten using search and replace patterns via rulesets. See Using rulesets.

• Use control sequence. See Control sequences.
• Enter phonetic input. See Inserting phonetic input.
• Tune with ActivePrompts. See Tuning TTS output with ActivePrompts.
• Specify pronunciations in user dictionaries. See Specifying pronunciations with user dictionaries.

When developing applications with synthesized prompts, listen to samples to identify areas that need tuning. There are several ways to listen to samples. Suggestions:

• Load a list of prompts into Nuance Vocalizer Studio and step through the prompts one by one.
• Load the items as a .csv file in Excel and create a Visual Basic script hook to the Vocalizer command line tool nvscmdline.exe. This method organizes the items neatly, and you can listen to each one by pressing the down arrow. The Excel sheet also facilitates annotation.