I've got about 300 DVD's, and perhaps 100 CD's, choc-full of mostly media stuff, whose contents I keep track of using the application GWhere. In spite of the fact that this app was last updated in 2007, it is quite good at what it does, it's only real problem being that it doesn't register a new DVD being inserted, and so has to be exited & restarted before it will read its contents.
So, everything was rosy on that front, until recently. I had noticed with Ubuntu 10.04, that whenever I ran GWhere, sometimes, very noticable graphics-corruption would occur surrounding the text-font it used. Although annoying, this wasn't a show-stopper until recently. With Mint 17.3, this now results in the OS entirely locking up, something that is very, very hard to achieve with Linux.
As this is completely unacceptable, alternative solutions were tried. First, as I suspect that this has something to do with an old Gtk graphics library being used, I tried building the package from source, in the hope that it would sort things. This proved to be a non-starter however, as the new(er) Gtk libraries are no longer compatible. Next I tried running the Win32 version under Wine, which worked to a fashion. Unfortunately, whenever you move the mouse-pointer over a Gtk-control, the mouse-pointer disappears! Assigned keyboard shortcuts seem to work, but this is not an acceptable solution either. Finally, the only thing that worked properly was to install the Win32 version on a VirtualBox'ed Win-XP and run it from there - something I'm not at all happy with, but beggars can't be choosers!
Another disk-cataloguer that I had heard good things about, and development-wise, had only been abandoned in 2014, was CdCat. This also had options to 'import' the contents of other Cataloguer softwares, which I had hoped would work with my already-existing GWhere database files. No such luck! Although it can import from 11 different Cataloguer applications, GWhere isn't one of them - which is stupid, given that as there were only 2 main contenders on the Linux scene in the first place (GWhere & CdCat) you would expect the CdCat developer to specifically cater for GWhere files. Or rather, it doesn't import the data in a usable form. The only import option that works is the really old (circa 2003) Linux 'Gtktalog' format. After extracting the compressed database file using 7zip, and adding a 'csv' extension, all the data is imported, but no directory structure is retained - rendering it useless! Incidentally, GWhere also has 'Import/Export' menu options, but unfortunately they are just there for show - what you have in reality are just a 'hook' entry-point for import/export plugins, something that was never developed...
Anyway, given that most of the available versions of CdCat that are built, don't use the most recent source - Ubuntu's 'Trusty' repository for instance has a positively ancient version 1.8, in contrast with the 'newest' CdCat 2.3.1 source - and particularly since its web-page lists the huge number of bugs that were removed in the final versions, I decided to build it from scratch. Not a trivial matter either, as it turned out. Some libraries are no longer compatible, some are completely missing. I eventually found all of its various dependencies and managed to build the package, which can be downloaded from here by anyone so inclined.
It was only after going through all of this, that I discovered that CdCat could not import GWhere's databases - WHICH SUCKED!!! I also then googled an old post of mine on the Ubuntu forum, something I'd completely forgotten about, where I criticised CdCat's speed at reading DVD's - up to a minute to read a disk, which is ridiculous! I haven't tried the newly-built version, but I've little doubt that it's the same. GWhere in comparison, takes 1-3 seconds.
So, I either stick with my present convoluted GWhere setup, or I laboriously re-create all of the databases from scratch using CdCat. Hmmm, or maybe start looking for something Windows-based...
24 May 2016
NaturalReader 14 text-to-speech application rant...
I have been a long-time user of NaturalReader, initially having bought version 6.6, which over time, I managed to upgrade online, up to version 9 in fact, before being informed that I was no longer entitled to any more free upgrades. Being also a long-time Linux user, I loathe paying for software, especially where there is a perfectly adequate free alternative available. Well, there wasn't in this instance, on either Windows or Linux - hence, the purchase, along the lines of $40 if memory serves!
In my opinion, the speech-engine used by NaturalReader is by far the best available, and which is licensed by NaturalReader from the mega-corp, AT&T. Although the Google speech-engine is very good, IMO, it still falls quite a bit behind its AT&T counterpart, particularly where higher speech-rates are concerned, though this 'edge' comes at a cost. AT&T's approach uses very large libraries of speech samples, or the phonemes-parts thereof, whereas Google, judging by its speech-engine footprint, uses a different technique, and though smaller, is less effective. So, although the AT&T (henceforth NaturalReader, or 'NR') approach necessitates the use of more disk-space/processing-time, the end-results are I think, worth the effort.
Just recently, out of curiosity more than anything else, I searched out a demo of the latest-and-greatest (sic) NR online, deciding to give it a spin. What immediately caught my eye was the installation size, a whopping 200meg - version 6.6 was about 25meg, whereas version 9 comes in at just under 50meg. So I figured that all those extra megs must be doing something interesting, and sure enough, NR now comes with a built-in Optical Character Recognition (OCR) feature, which piqued my interest still further.
Given that I'm stuck with using long-defunct VirtualBox'ed Windows XP's (Win7 being both too slow and taking up too much disk-space) I thought this would prove to be a problem. Imagine my surprise when it installed without a hitch. Then imagine my disgust at finding that it would not run without NET 3.0 having to be installed! After downloading and installing this, re-running NR saw it spring to life. Or maybe 'lurch' is more apt - it's painfully slow to load, and more so in operation! But imagine my disgust at finding that this 200meg behemoth doesn't include any proper voices - you are stuck with using Microsoft's Sam, which is, ahem, less than adequate. So off I go asearching again, this time for a disk containing my 'Mike-US' voice for NR 9, all 700megs worth. After installing this, I was relieved to find that it was detected and worked without a problem, albeit as I mentioned, a mite sluggishly.
Next up, a check of NR's OCR prowess. Or at least that was the intention. In practice, it crashed every time I tried it! So, the main reason for installing it in the first place proved to be its epic-fail. Sucks huge.
But my main reason for this blog entry is to point out one simple fact. Although NaturalReader has 'progressed', at least version-wise, from 6.6 to 14 on my watch, the speech quality of the files it produces has not improved one iota! What you are getting now for your money ($120 for the 'Pro' version) is a piece of bloatware that NaturalReader keeps adding bits to in order to justify the version (and price) increases. Versions 6 through to 9 were definite lack-lustre affairs. These softwares were so badly designed, that, for example, you could not search & edit a piece of text simultaneously - you needed to close the Search dialogue prior to editing text due to the shoddy 'modal' nature of the Search menu employed. This was barely acceptable with Windows 3.1, never mind Win-XP and beyond. NaturalReader 14 fares little better in my opinion. Huge file size, dependent on NET, therefore slow as hell, topped off with crashes at the bits that may be of interest.
Do yourself a favour and leave this dog alone. Granted, NaturalReader produces some of the best text-to-speech output there is, but kudos must go to AT&T for this, not the still-lame attempt by NR to 'package this TTS-engine proficiently. If you must go with NaturalReader, search out version 9. Primitive as hell, feature-wise, but it's fast, and sound quality is still the same as with version 14, and it's NR's best attempt, before it became completely dependent upon M$ NET.
In my opinion, the speech-engine used by NaturalReader is by far the best available, and which is licensed by NaturalReader from the mega-corp, AT&T. Although the Google speech-engine is very good, IMO, it still falls quite a bit behind its AT&T counterpart, particularly where higher speech-rates are concerned, though this 'edge' comes at a cost. AT&T's approach uses very large libraries of speech samples, or the phonemes-parts thereof, whereas Google, judging by its speech-engine footprint, uses a different technique, and though smaller, is less effective. So, although the AT&T (henceforth NaturalReader, or 'NR') approach necessitates the use of more disk-space/processing-time, the end-results are I think, worth the effort.
Just recently, out of curiosity more than anything else, I searched out a demo of the latest-and-greatest (sic) NR online, deciding to give it a spin. What immediately caught my eye was the installation size, a whopping 200meg - version 6.6 was about 25meg, whereas version 9 comes in at just under 50meg. So I figured that all those extra megs must be doing something interesting, and sure enough, NR now comes with a built-in Optical Character Recognition (OCR) feature, which piqued my interest still further.
Given that I'm stuck with using long-defunct VirtualBox'ed Windows XP's (Win7 being both too slow and taking up too much disk-space) I thought this would prove to be a problem. Imagine my surprise when it installed without a hitch. Then imagine my disgust at finding that it would not run without NET 3.0 having to be installed! After downloading and installing this, re-running NR saw it spring to life. Or maybe 'lurch' is more apt - it's painfully slow to load, and more so in operation! But imagine my disgust at finding that this 200meg behemoth doesn't include any proper voices - you are stuck with using Microsoft's Sam, which is, ahem, less than adequate. So off I go asearching again, this time for a disk containing my 'Mike-US' voice for NR 9, all 700megs worth. After installing this, I was relieved to find that it was detected and worked without a problem, albeit as I mentioned, a mite sluggishly.
Next up, a check of NR's OCR prowess. Or at least that was the intention. In practice, it crashed every time I tried it! So, the main reason for installing it in the first place proved to be its epic-fail. Sucks huge.
But my main reason for this blog entry is to point out one simple fact. Although NaturalReader has 'progressed', at least version-wise, from 6.6 to 14 on my watch, the speech quality of the files it produces has not improved one iota! What you are getting now for your money ($120 for the 'Pro' version) is a piece of bloatware that NaturalReader keeps adding bits to in order to justify the version (and price) increases. Versions 6 through to 9 were definite lack-lustre affairs. These softwares were so badly designed, that, for example, you could not search & edit a piece of text simultaneously - you needed to close the Search dialogue prior to editing text due to the shoddy 'modal' nature of the Search menu employed. This was barely acceptable with Windows 3.1, never mind Win-XP and beyond. NaturalReader 14 fares little better in my opinion. Huge file size, dependent on NET, therefore slow as hell, topped off with crashes at the bits that may be of interest.
Do yourself a favour and leave this dog alone. Granted, NaturalReader produces some of the best text-to-speech output there is, but kudos must go to AT&T for this, not the still-lame attempt by NR to 'package this TTS-engine proficiently. If you must go with NaturalReader, search out version 9. Primitive as hell, feature-wise, but it's fast, and sound quality is still the same as with version 14, and it's NR's best attempt, before it became completely dependent upon M$ NET.
Subscribe to:
Posts (Atom)