SvarDOS community forum

Over at Mastodon someone asked about BNS support for DOS. I want to recommend the person to try out the BNS build. How is its state? Anyone tested / used it recently? I am not into the topic, so I wanted to ask here before recommending it...

This is what Georgina Brummell wrote to me: https://someplace.social/@dandylover1/115204737870929908 Not sure the above link works. So I copy it here: @boeckmann *Most of this will probably make no sense to you, so I will try to explain it briefly. But this is good to keep in mind when you speak with the other developers.* That is very interesting! I have never heard of this version of DOS and will need to research it. Provox, to my knowledge, is the only fully open source screen reader available, though I have never used it myself. I am trying to locate Larry Skutchan, the creator of another screen reader, called ASAP, to see if he will allow that to be open source as well, since it's very flexible and development continued well into the 1990's. Regardless, the fact that this is even included at all is a great first step! Please ask if anyone is actively developing Provox or if it is just set up so that it comes with the system itself. One of the main problems with older screen readers is that they require a hardware synthesizer connected to a serial port, or a simulated one if in a virtual machine and using a Windows screen reader such as NVDA. There were tones for DOS, such as JAWS and Flipper, that used the speech in SoundBlaster, but that required very specific hardware, namely the soundcard. My dream is to be able to integrate something such as ESpeak into DOS, so that a screen reader can work with software as is done in Windows.

The mentioned ESpeak seems to have traces of DOS support (DJGPP) in its source. It is however neither in the SvarDOS nor in the FreeDOS repo. Wondering if we can get it to compile. Will try it in the next few days... https://sourceforge.net/projects/espeak/files/

As far as I can tell it works. But it is nothing big, really. There is no provox development involved, the BNS build is basically the usual SvarDOS but with Provox pre-installed. I have also tuned the SvarDOS installer so it is somewhat "screen-reader-friendly". I was able to install SvarDOS on my 386 without using eyes. Not a pleasant experience, but possible. Once the system is installed, it is no different than any other DOS that would have provox installed. And yes, a hardware BNS connected to COM1 is very much required (I don't have one, so I wrote the emuBNS emulator and ran it on a RaspberryPI with a USB-to-COM adapter). Mateusz

> The mentioned ESpeak seems to have traces of DOS support (DJGPP) in its source.

Compiling espeak is not enough. Georgiana would like a screen reader that talks on its own, ie without TTS hardware. To achieve this, a TSR screen reader like Provox. ASAP, JAWS, etc would have to output speech instructions to espeak, and the latter would have to generate samples in realtime and feed them to a Sound Blaster. All this while the actual application is being operated by the user. Not a trivial feat. Mateusz

Thanks, I now understand better how this is supposed to work. ESpeak than seems to be barely useful, I guess, without having any output capability other than outputting .WAV files and phonemes for mbrola. I would not say that it is impossible to implement, but having it run in the background as some kind of TSR (it is a 32-bit DJGPP executable!) in addition to provox, with the user application running "as normal", seems indeed a rather complex (to say at least) task. I will reply to Georgiana that this probably is a dead end. Bernd

> I will reply to Georgiana that this probably is a dead end.

Nothing is impossible, but the amount of required work to achieve a functional ScreenReader-with-TTS-output TSR is certainly huge. Probably unrealistic on anything less than 32bits, but on some pentium with lots of RAM and a sound card dedicated to TTS, it should be possible, either as an oldschool TSR or as a JEMMEX extension - if only there is someone out there willing to dedicate a few months of his life creating such thing. For any practical matter I think the only realistic path is to use an external synth, like a Braille 'n Speak. These things are rare and tend to be expensive nowadays, that's why I had created an emulated version that can be run on a cheap RPi (vide my Rasp 'n Speak system). https://emubns.sourceforge.net/ https://emubns.sourceforge.net/raspnspeak.html Mateusz