Google launches Voice Search

Background

Google's speech work began inside the company in 2005, when it formed a speech group ^[4]. That group's first public product was GOOG-411, launched in 2006: a directory assistance service driven by speech recognition ^[4]. Before voice input reached general web search, Google had already put speech into maps. In March 2008 it introduced its first multi-modal speech application, for Google Maps for Mobile ^[2]. Multi-modal here meant that speaking was one of several ways to enter a query, alongside tapping and typing, and the target was a specific kind of result: local businesses on a map ^[2].

What happened

On 17 November 2008, Google announced voice search in the Google Mobile App for iPhone ^[1]. The interaction was built around the way people already hold a phone: the user holds the phone to their ear, waits for a beep and speaks the query, without tapping any buttons ^[1]. Google said the feature was enabled by default for U.S. English users only ^[1].

Google Research later described the same release as the point where voice search stopped being a maps-only feature and became whole-web search: the November 2008 Google Mobile App for iPhone included a search-by-voice feature, extending voice search from searching for businesses on maps to searching the whole web ^[2]. A Google Research paper on mobile speech recognition records the launch in broader terms, stating that Google launched Voice Search in the US in 2008 for several types of smartphones ^[4].

Key details

Recognition did not happen on the handset. Audio was sent to Google's servers, recognized there, and both the recognition and the search results were returned to the phone ^[4]. Splitting the work this way means the phone only has to capture and upload sound, while the heavy computation runs on machines the service provider controls; from the user's side, spoken words are turned into text and then into results, with the middle step invisible.

The design also set the boundaries of the launch. The feature was on by default for U.S. English users, not for everyone ^[1]. The default mattered because it decided who could try voice search without changing any settings.

Why it mattered

Google's own retrospective counts voice search among its 2008 milestones, listing it as introduced in the Google Mobile App and extended to desktop computers in 2011 ^[3]. Reading the dates together shows a clear path. GOOG-411 in 2006 handled one narrow task, directory assistance ^[4]. The March 2008 application for Google Maps for Mobile kept speech inside maps and inside local business queries ^[2]. The November 2008 Google Mobile App opened spoken queries to the whole web ^[2], and the 2008 Voice Search launch put the capability on several types of smartphones in the US ^[4]. Step by step, the range of questions that could be spoken widened, and speech moved from an aid in one specific situation toward an ordinary way of filling in a search box.

The limits were just as visible. At launch the default-on feature covered U.S. English users only ^[1], so the people who could use voice search out of the box were a narrowly defined group, and everything beyond that group came later.