25th May 2010

Speeka

Created initially as a Word Press plugin in 2008, Speeka also supports conversion of generic html pages to MP3.

Once a page is registered with the system, Speeka’s crawler regularly checks the page content for changes. Speeka’s capacity is reduced these days. Though able to serve existing users without issues, it’s just running from a single dedicated server now largely for closed project use only.

I’ve left this page up as there aren’t many places where you can compare the various open source speech engines.

There are now several commercial alternatives available which are simple to implement for web – eg Voice Forge, ReadSpeaker, BrowseAloud, ROK Talk and more than a few plugins for WordPress have emerged.

For developers The Web Speech API is being widely adopted by browsers and provides efficient text-to-speech with relative ease.

Text-to-Speech Demos

To quickly compare voices use the pre-rendered clips beneath the form.

Use the following to test the supported voices and parsing – text or html input is acceptable. Optional post processing through Sox can improve some of the harsher voices.

Input is annotated basically by default – headings are noted and acronyms and abbreviations are expanded. Additional elements such as forms, lists, addresses, tables and links are announced if the semantic annotation checkbox is ticked. Forms and tables are not reliably processed, their presence is noted to the listener but content is simply linearised.

The audio’s ID3 content is constructed from page title, the encoded text content and user name.

Pre-rendered Voices for Quick Comparison

Young British Male – Mbrola EN1
Play voice_en1.mp3
Old British Male – CSTR ‘rab’ diphone
Play voice_rab.mp3
Scottish/UK Male – CMU ‘awb’ arctic hts
Play voice_awb.mp3
Scottish/UK Male – Nitech ‘awb’ arctic hts
Play voice_awb2.mp3
Scottish/US Male – CSTR ‘ked’ diphone
Play voice_ked.mp3
Thoughtful US Female – CMU ‘clb’ arctic cg
Play voice_clb.mp3
Thoughtful US Female – Nitech ‘clb’ arctic hts
Play voice_clb2.mp3
Bored US Female – Mbrola US1
Play voice_us1.mp3
Curious US Female – CMU ‘slt’ arctic cg
Play voice_slt.mp3
Curious US Female – Nitech ‘slt’ arctic hts
Play voice_slt2.mp3
Bored US Male – Mbrola US2
Play voice_us2.mp3
Smooth US Male – Mbrola US3
Play voice_us3.mp3
Lazy US Male – CSTR ‘kal’ diphone
Play voice_kal.mp3
Chilled US/Canadian Male – CMU ‘jmk’ arctic cg
Play voice_jmk.mp3
Chilled US/Canadian Male – Nitech ‘jmk’ arctic hts
Play voice_jmk2.mp3
Narrator US Male – CMU ‘rms’ arctic cg
Play voice_rms.mp3
Narrator US Male – Nitech ‘rms’ arctic hts
Play voice_rms2.mp3
Precise US Male – CMU ‘bdl’ arctic cg
Play voice_bdl.mp3
Precise US Male – Nitech ‘bdl’ arctic hts
Play voice_bdl2.mp3
Lancashire Male – eSpeak en-n
Play voice_en-n.mp3
Standard English Male – eSpeak en
Play voice_en-def.mp3
RP Male – eSpeak en-rp
Play voice_en-rp.mp3
Brummie Male – eSpeak en-wm
Play voice_en-wm.mp3
West Indian Male – eSpeak en-wi
Play voice_en-wi.mp3
Scottish Male – eSpeak en-sc
Play voice_en-sc.mp3
US Male – eSpeak en-us
Play voice_en-us.mp3
Esperanto – eSpeak eo
Play voice_eo.mp3
French – eSpeak fr
Play voice_fr.mp3
German – eSpeak de
Play voice_de.mp3
Russian – eSpeak ru
Play voice_ru.mp3
Polish – eSpeak pl
Play voice_pl.mp3
Italian – eSpeak it
Play voice_it.mp3
Spanish – eSpeak es
Play voice_es.mp3
Latin – eSpeak la
Play voice_la.mp3
Commercial Voices for Contrast
British Male – Cepstral Lawrence
Play voice_Lawrence.mp3
British Female – Cepstral Millie
Play voice_Millie.mp3
US Male – Cepstral David
Play voice_David.mp3
British Male 2 – Loquendo Simon
Play voice_LoqSi.mp3
British Female 2 – Loquendo Kate
Play voice_LoqKa.mp3
US Male 1 – Microsoft Sam
Play voice_MSSam.mp3
US Male 2 – Microsoft Mike
Play voice_MSMike.mp3
US Female – Microsoft Mary
Play voice_MSMary.mp3