Participants for voice recognition testing

What we want and why

Voice Trax West has projects several times a year where we bring in non-professionals to record short spoken commands and sentences to test new devices for electronics companies. These recordings are used to improve the speech recognition technology for specific products.

For example, if an alarm clock will respond when people say “What time is it?” we would collect recordings of various people speaking that particular phrase. Those recordings will be used to test and improve the speech recognition software, so the alarm clock can better understand a wide variety of voices.

Who should record

No experience is necessary to join our list, but if you have some experience that is definitely ok too.

We want native speakers of the specified language, but we're not looking for professional voice talents; we want real people using their normal voices, and the full array of typical accents. Because we use these recordings to test and improve our technology, the speech should be as representative of the target demographic as possible. For example, for “US English”, we would want speakers with accents from all different regions, such as West coast, New England, Southern states, and midwest.

We will often need to record as many as 100 people or more.

What do we mean by regions

It is common that we might move around the United States, so it is NOT crucial that an accent is pure. That would make things impossible to capture perfectly. When looking at what region to label yourself as, a good rule of thumb is to use where you grew up in your early years from birth through six to ten years of age. If it is more complex than that then just choose what you think best fits.

We are typically required to have an even split between these regions to the best of our ability, so it is important that we have that information.

Does it pay?

YES. We do pay you for your time. Tyically it is in the range of $30 for a half hour or less, or $60 for closer to an hour or less.

Definitions of Regions

We want native speakers of the specified language, but we're not looking for professional voice talents necessarily; we want real people using their normal voices, and the full array of typical accents. Because we use these recordings to test and improve voice recognition, the speech should be as representative of the target demographic as possible. For example, for “US English”, we would want speakers with accents from all different regions, such as West coast, New England, Southern states, and Midwest.

West (California, Oregon, Washington, Arizona, Nevada, Utah, Idaho, Wyoming, Colorado)

Inland North (Chicago, Milwaukee, southern Michigan, western NY)

North Central (North Dakota, northern Minnesota, Michigan upper peninsula)

North (South Dakota, Iowa, southern Minnesota, northern Wisconsin)

Mid-Atlantic (DC and surroundings)

Midland (Ohio, Indiana, Missouri, Kansas)

St. Louis Corridor (between St. Louis and Chicago)

West Pennsylvania (Pittsburgh)

West New England (Albany, Vermont, Western MA, Western CT)

East New England (*Boston*, NH, Eastern MA, Eastern CT)

NYC (Preferably NYC, as opposed to Long Island)

South (Southern Texas, Louisiana, Arkansas, Mississippi, Georgia, South and North Carolina, Southern Virginia, West Virginia)

Texas South (West Texas, maybe Dallas)

Inland South (Birmingham, AL, eastern TN, western NC and SC)

US English Signup

US Spanish Signup