Espeak, Flite and MaryTTS
 

2015-12-12

I wonder if anyone has done a comparison among eSpeak, Flite and MaryTTS in terms of their suitabilty to use in Linux scripts.

I consider Espeak and Flite extremely simple and easy to use in my show-and-tell bash scripts. Even though I've only a nodding acquaintance with MaryTTS, I find its extreme slowness alone
unacceptable even though I rate its sound quality above that of Flite and eSpeak.

As for pronunciation correctness, I find Flite superior to eSpeak, which mispronounces words like squirrel and fan. On the other hand, eSpeak always sounds very clear and sufficiently loud whereas Flite's voices all seem to sound muffled and sometimes quiver.

I feel that my best compromise would be to stay with Flite and hope that somehow I can render its sound clearer and louder.

I'll appreciate very much any help and advice. I wouldn't mind paying a reasonable sum for a commercial Linux TTS that is free from the above applications' basic drawbacks.

Julianvb

what do you mean "suitabilty to use in Linux scripts"?
iirc, all of them are command line utilities, so why shouldn't it be possible to use them in scripts?
as i already pointed out, i find flite much better than espeak.
flite can read from a file or by directly inputting a string.

about volume:
i can't hear much difference in loudness.
are you sure you are using the voices?

Hi, ondoho,

It's true that all TTS engines can be used with scripts. My question is how much work the user wants to put in to render one such as MaryTTS, workable. MaryTTS takes several seconds to react to a maryspeak command even on the command line. I do not wish to spend weeks making a script work. As far as implementation goes, I find eSpeak and Flite a cinch and a pleasure to use. It's true that both eSpeak and Flite can receive text inputs from a file, the command line, or within a script. I've been experimenting with TTS scripts for several years, mostly to entertain my soon-to-be-three grandchildren bilingually.

I have tried all the available Flite voices: awb, rms, slt, cbl, kal, awb16 and kal16 in scripts or on the command line. They all sound muffled, remote and weak. Spontantaneity in very essential to my scripts.

Perhaps this latter problem is hardware-related. My 2004 Compaq desktop uses a pair of Gateway speakers. Flite's voices don't change much when I try an external speaker with my 2013 Dell laptop. I've consulted Alan W. Black, author of Flite about my sound problem with Flite. He suggested that I might be able to improve the acoustic quality by using some old loudspeakers.

Can you tell me what computers you are using? Thanks very much for your response.

Julianvb

just some external creative speakers with my desktop.

there's 16 .flitevox files for usa english. tried all of them?

i just tried to pipe flitevox' output to play (sox), but it doesn't seem to be possible.
you should ask the developer if it's possible to use standard output, so as to pipe the output to another utility.