E-Speech offers letter-to-phoneme software that provides the most accurate name pronunciation available today for the names of people, businesses, and places in the United States. E-Speech software is also excellent at pronouncing English words, so that it can be used for general text as well as for name-oriented applications.

Our software is based on intelligent letter-to-phoneme rules, but it also includes verified pronuniciations for many thousands of names. It can be used with speech recognition and with text-to-speech, and it supports all popular phonetic alphabets (IPA, ARPAbet, CPA, SAMPA, HARK, or "Newspaper" phonetics).

The software translates input in English orthography into a phonological representation, indicating phonemes, word-stress, syllable boundaries, word and phrase boundaries, and optionally, intonation and phrase-stress.

In addition, our software provides multiple pronunciations for many names that have more than one possible pronunciation. For example, "Weinstein" can be pronounced as "wine STINE" or "wine STEEN," "Migliore" may be pronounced as "mig lee OR", "mig lee OR ee", "mig lee OR ay", "mil YORE", "mil YORE ee", or "mil YORE ay," and "Cai" can be pronounced as "tsy" or "chy" or "sy". This feature is important for speech recognition applications, since they need to link all the pronunciations to the same name.

Our letter-to-phoneme software can be used real-time during speech recognition/synthesis transaction processing or off-line to build pronunciation dictionaries that can be incorporated into your ASR or synthesis software. The software is available as part of the E-Speech synthesizer or as a standalone system. It runs on Win95/98/NT/2000/XP/Vista and various Unix systems, including Linux, and is easy to integrate with applications. The standalone version is available as a command-line executable or as a library/DLL containing C-callable functions. We welcome customization: we are happy to tailor our software to your specific needs.