In this press release, automated speech processing specialist SVOX AG reveals interesting details about the mix of “voices” that have been downloaded from the Android marketplace to support navigation applications, ebook reading, game playing, speech-to-speech translation and other talking apps. While it is unclear how much control end-users (as opposed to application developers) have in choosing the voice of their device, the mix of languages and gender selection is revelatory.

In over 80% of instances, mobile users download a female voice for their device or application. For English speakers, the figure is closer to 85%. By the way, the combination of US and UK English accounts for more than half of all downloads, with a cluster of Russian, French and German accounting for something on the order of 6% each.

The predominance of English-speaking applications reflects the relative distribution of smartphones in general (and Android in particular) as well as the mix of applications that are speech-enabled. The preference for female voices is a bit of a reversal for the speech processing industry. Mobile ASR (automated speech recognition) is notoriously biased against female voices. At first it was written off as the product of grammar development built primarily on male utterances. Later there was a mismatch between the range and pitch of female voices and the sweet spot in the phone lines’ (and wireless links’) information carrying capabilities.

This gender gap may be short-lived or, ultimately, of little meaning. Choice of preferred voice is either application driven (most navigational applications and devices use female voices) or developer driven – meaning it is out of the hands of the end-user. The next generation of “life-like” TTS is designed to be extremely flexible to the point where “publishers” or entertainment content providers can assign specific voices to individual characters. In these cases, the mix of gender will move to reflect the general population – something close to 50/50.

Categories: Articles

Tags: speech processing, Mobile Speech Apps, Text-to-Speech, SVOX