How big is the speech recognition market?

There is no single size. The published worldwide figure for speech and voice recognition has a median of 21.66 billion US dollars ¹, but that median mixes a current-year estimate with a forecast. The estimates we track come from 12 sources ² and total 32 data points ³. Each describes a different market, region or year, so the spread says more about definitions than about the industry.

The only worldwide estimate for this year is from one firm

MarketsandMarkets puts the global speech and voice recognition market at 9.66 billion US dollars for the report's own year ⁴. It projects the same market to reach 23.11 billion US dollars by the end of its forecast window ⁵. That is the range the median above comes from: one firm's estimate and the same firm's forecast.

Both numbers are a firm's estimate, not a measurement. The release does not publish a method, and the scope is broader than speech recognition software alone. Treat them as one analyst's view, not as a consensus.

Narrower definitions give much smaller numbers

P&S Intelligence counts only speech recognition software, and only in the United States. It puts that market at 4.4 billion US dollars for the report's year ⁶, up from 3.9 billion the year before ⁷. Its forecast reaches 10.1 billion US dollars by the end of its window ⁸.

You cannot subtract this from the worldwide figure to find "the rest of the world". The two firms draw their boundaries differently, and neither publishes enough on the page to reconcile them.

Broader definitions give much larger numbers

Two estimates are far above the rest because they measure something else.

• Transcription services. Grand View Research values the U.S. transcription market at 30.42 billion US dollars ⁹ and forecasts 41.93 billion by the end of its window ¹⁰. This is a services market. It includes the work people do, not only software. The median for the two transcription figures is 19.2 billion US dollars ¹¹.
• Voice assistants. Roots Analysis estimates the voice assistant application market at 8.1 billion US dollars for a past year ¹² and projects 153.5 billion a decade later ¹³. Its growth rate is far above the other firms' rates, and it comes from a low starting base. The page's own attributes table gives a slightly different end value, a small internal inconsistency.

The voice assistant median of 80.8 billion US dollars ¹⁴ sits between a current-year figure and a ten-year forecast, so it describes neither.

How the estimates line up

The first chart shows estimates for the same years, and the second shows how far apart they are.

Showing 8 of 13 rows. See every row in the dataset.

What these numbers do and do not measure

• They are estimates by research firms. Most pages do not describe their method, so you cannot check how a figure was built.
• Forecasts are not measurements. A forecast depends on the firm's own growth assumption. Compare forecasts only with forecasts for the same year.
• Scope differs. The dataset mixes worldwide and U.S.-only figures, software and services, and speech recognition and voice assistants. Only figures for the same market, region and year are comparable.
• Coverage is thin. With 12 sources, a single firm's choice can move a median a long way. Nothing here counts users or devices.
• Only publishers' own pages count. Figures repeated by blogs and aggregators are left out.

If you quote a market size, name the firm, the market, the region and the year. A bare "billion dollars" figure without those four is not meaningful.

What this means if you are choosing how to dictate

Market size does not tell you which tool is better. A large market means many vendors, but it says nothing about accuracy, privacy or price for your use. To pick a tool, compare products directly on accuracy, languages, where audio is processed and cost. Our market comparison does that, and the guide to the speech-to-text API explains what developers buy when they build recognition into their own products.

For everyday writing, the free tools built into your computer or phone are often enough. If you want on-device dictation for Mac and Windows, Yap, which we make, is one option to compare. See dictation software for how the main kinds differ.

For the underlying technology, read our overview of speech recognition. To see every estimate with its publisher, region and year, open the full dataset.