Swift

From LQWiki
Jump to navigation Jump to search

Introduction

Swift is a text-to-speech engine made by Cepstral allowing the user to convert text to computer generated speech. For general use, Swift's functions are comparable to that of Festival and other free solutions, allowing one to easily convert text to speech, and the ability to save that speech as an audio file (e.g. .wav).

One might find that the most notable pro to using this commercial product, compared to a free solution, is the quality of the voices available. (use the external links below to compare two systems). Cepstral offers 8KHz and 16KHz voices at a reasonable cost (when compared with other non-free solutions) and are compatible with the Festival project (see the Festival script on the Cepstral Support Pages.

One can download and test voices without cost, however there is a Cepstral demo notice at the beginning of each sentence.

How to install

Cepstral is not available readily from package repositories. You need to download it and install it via http://www.cepstral.com.

gunzip filename.tar.gz && tar xf filename.tar
  • or, in most distribution's you can right click the file and 'extract to folder'
  • Go to the new folder and run:
./install.sh
  • Follow the prompts

How to register a voice

swift --reg-voice
  • Follow the instructions and enter your name and product key as indicated on the invoice.

How to use Cepstral

Once installed, you can use swift similarly to Festival:

$ swift "Hello, this is a test."

The following uses the Allison voice, to convert a raw text file to a wave file, not unlike Festival text2wave:

$ swift -n Allison -m text -f /path/to/file.txt -o file.wav

See also

External Links