MobileRead Forums - View Single Post - KSpeak

notimp · 01-30-2017, 08:27 AM

@encol: No criticism against this project or you personally - between "what Amazon has chosen to provide on the device" and there being no TTS support at all - its still great that it exists.

Also - many of the shortcomings are more structural issues than your fault.

- Open source TTS engines not being up to par with commercial solutions > the "big five" tech giants use big data and machine learning to "understand and improve" their TTS engines - you can't do that with "five people in a garage" any longer... (I think its worthy to think about it from that perspective as well, as "voice assistents" are currently held up as "the modern user interface" - so when user expectations (not mine) go in one direction, and open source projects cant - there is a structural reason for it).

- Amazons "dongle" implementation is limited, restrictive, follows their own user experience deliberations and probably isn't "standards oriented", so when a certain popup crashes a process that isn't supposed to run in the first place - thats hardly your fault either.

- Leaving the dongle in while the device auto reboots causing a bootloop - is an unforeseen circumstance (in conjunction with the mobileread recovery scripts) and an issue thats probably impossible to solve, because you are weighing the need for a hack to allow people with bricked devices to enter recovery in a way thats low level enough that a screen input is not needed.

- And a part of people that would usually wan't to use a TTS engine can't be expected to be able to troubleshoot all this, because they might be visually impaired.
-

I'm interested in checking out new TTS implementations, when they come along - but mainly from that perspective (how "usable are they" for me and other folks). So thats basically me saying - valiant effort, good job - great for even venturing into this field and creating something that works - but there are limits, that others should know exist -

in this implementation, and in others on this device.

IVONA on Android is also free (but you have to jump through more and more hoops to get it set up (because of Amazons decision to get voices depublished - because of FBReader resorting to a "more simple UI experience", that doesn't allow you to change the document language on the fly - in newer versions (which is important for TTS purposes) ), and might be a better solution if you are really looking for one, because you need it, rather than, because its cool to play with one.

Its cool to play with KSpeak as well.

01-30-2017, 08:27 AM	#33
notimp Addict Posts: 248 Karma: 892441 Join Date: Jul 2010 Device: K2i	@encol: No criticism against this project or you personally - between "what Amazon has chosen to provide on the device" and there being no TTS support at all - its still great that it exists. Also - many of the shortcomings are more structural issues than your fault. - Open source TTS engines not being up to par with commercial solutions > the "big five" tech giants use big data and machine learning to "understand and improve" their TTS engines - you can't do that with "five people in a garage" any longer... (I think its worthy to think about it from that perspective as well, as "voice assistents" are currently held up as "the modern user interface" - so when user expectations (not mine) go in one direction, and open source projects cant - there is a structural reason for it). - Amazons "dongle" implementation is limited, restrictive, follows their own user experience deliberations and probably isn't "standards oriented", so when a certain popup crashes a process that isn't supposed to run in the first place - thats hardly your fault either. - Leaving the dongle in while the device auto reboots causing a bootloop - is an unforeseen circumstance (in conjunction with the mobileread recovery scripts) and an issue thats probably impossible to solve, because you are weighing the need for a hack to allow people with bricked devices to enter recovery in a way thats low level enough that a screen input is not needed. - And a part of people that would usually wan't to use a TTS engine can't be expected to be able to troubleshoot all this, because they might be visually impaired. - I'm interested in checking out new TTS implementations, when they come along - but mainly from that perspective (how "usable are they" for me and other folks). So thats basically me saying - valiant effort, good job - great for even venturing into this field and creating something that works - but there are limits, that others should know exist - in this implementation, and in others on this device. IVONA on Android is also free (but you have to jump through more and more hoops to get it set up (because of Amazons decision to get voices depublished - because of FBReader resorting to a "more simple UI experience", that doesn't allow you to change the document language on the fly - in newer versions (which is important for TTS purposes) ), and might be a better solution if you are really looking for one, because you need it, rather than, because its cool to play with one. Its cool to play with KSpeak as well. Last edited by notimp; 01-30-2017 at 08:32 AM.