Release EDDI 3.3 - Bring your cockpit to life

[T'kael]

yes, my friend, but it will not be useful, as my preferred choice for EDDI is Ivona Carla, and the new version discharges it as "not usable".
Also, i've US and UK MS voices installed, and they are all working in the new version, so i can't take this as an indicator that Elsa voice is broken.

Hi Parduz, it's difficult for me to offer a solution when I cannot replicate the problem. If you install EDDI and your voices on another PC then do you see the same problems?

The powershell method you mentioned works with "Microsoft Elsa Desktop" but breaks with "Microsoft Elsa". This is what we expect, since "Microsoft Elsa" uses the Windows Media synthesizer and "Microsoft Elsa Desktop" uses the System.Speech synthesizer (this is an important distinction so if you test with "Desktop" version then it is important to include this detail).

We prefer Windows Media voices when they are available, but for some unclear reason Microsoft Elsa breaks when your Windows Media synthesizer tries to use it. What I can do is try adding an additional test against the Windows Media voice. If that test fails then we would discard "Microsoft Elsa" and add "Microsoft Elsa Desktop" to the drop-down instead.

I'm not familiar with Balabolka but as a rule I do not trust closed source programs from sources that I do not know to be reputable (and there would be minimal benefit to trying because I wouldn't be able to compare code in any case).

 

[Parduz]

Not until you run Chkdsk /f and reboot

Done and done.
My SSD drive was perfect.
EDDI 4.0.3b2 still crashes when trying to use Elsa voice, and still doesn't lists the other non MS voices.

it's difficult for me to offer a solution when I cannot replicate the problem.

I bet

If you install EDDI and your voices on another PC then do you see the same problems?

I have no PC ready for this, but i can sort one if really needed: it will require a clean install of Win10 on an old (slow) laptop, and it will take a lot of time, so pls ask me to do it only if you're sure is necessary; i trust your judgement on this.

I'm not familiar with Balabolka but as a rule I do not trust closed source programs from sources that I do not know to be reputable (and there would be minimal benefit to trying because I wouldn't be able to compare code in any case).

It is just a relatively old and well known freeware text-to-speech.
I mentioned it just to say that also other programs are able to use my Ivona voice, other than the previous version od EDDI.

And this is just the thing i don't get: the "new test" in this new version doesn't want voices that are working flawlessy (i played 3 hours today usindg EDDI 4.0.2) and i don't get why they are so necessary that i need to check my pc and/or install stuffs here or on another PC when (from this side of the screen) those tests could be removed and studied with more ease until we understand what's the problem.

 

[Parduz]

We prefer Windows Media voices when they are available, but for some unclear reason Microsoft Elsa breaks when your Windows Media synthesizer tries to use it

Do this script uses Windows Media? If yes, then this script with Elsa works, on my PC

PS F:\- Games\== Elite Dangerous Stuffs ==> ./out-ssml-winrt.ps1 Prova Elsa


type  : sentence
time  : 1
value : Prova
start : 82
end   : 86

type  : word
time  : 101
value : Prova
start : 82
end   : 86

Do this helps?

 

[T'kael]

I have no PC ready for this, but i can sort one if really needed: it will require a clean install of Win10 on an old (slow) laptop, and it will take a lot of time, so pls ask me to do it only if you're sure is necessary; i trust your judgement on this.

Debugging issues in the code also takes many hours and is much easier if I know how to replicate the issue. As you already know, neither I nor @nepomuk have been able to replicate the issue exactly. If you were to replicate this on another PC then you could provide valuable information about how I could replicate the issue.

And this is just the thing i don't get: the "new test" in this new version doesn't want voices that are working flawlessy (i played 3 hours today usindg EDDI 4.0.2) and i don't get why they are so necessary that i need to check my pc and/or install stuffs here or on another PC when (from this side of the screen) those tests could be removed and studied with more ease until we understand what's the problem.

The new version of the file is extremely similar to the old file and I don't believe the problem with Microsoft Elsa comes from the added test. There are some minor differences to lock statements between the two versions.

I've found a way to replicate a similar error to the one you're seeing using those locking statements but when I use that method speech fails for all Windows.Media voices and not just for Microsoft Elsa.

I plan to revert the locking statements to match 4.0.2. Let's see if that resolves the issue with Microsoft Elsa. If not, the powershell script you listed above (which does indeed appear to use the Windows.Media synthesizer) looks like it may be a good starting point for any further debugging.

 

[T'kael]

yes, my friend, but it will not be useful, as my preferred choice for EDDI is Ivona Carla, and the new version discharges it as "not usable".

According to the log you submitted, EDDI didn't encounter Ivona Carla after it retrieved the available voices from the synthesizer. Could you please run a quick test for me and see if the voice is found when you run EDDI in standalone mode (without VoiceAttack)? If there is a difference then it might have to do with the fact that VoiceAttack runs as a 64 bit program while EDDI in standalone mode defaults to 32 bit.

Also, to be clear you are saying that Ivona Carla was found and worked with version 4.0.2?

 

[Parduz]

According to the log you submitted, EDDI didn't encounter Ivona Carla after it retrieved the available voices from the synthesizer.

oh! i missed this. I searched the log for "skipping" and overlooked the fact that Carla wasn't mentioned at all...

Could you please run a quick test for me and see if the voice is found when you run EDDI in standalone mode (without VoiceAttack)?

Never had VoiceAttack, i always use EDDI alone.

Also, to be clear you are saying that Ivona Carla was found and worked with version 4.0.2?

ABSOLUTLY sure.
It is the default voice, everything is said with that voice (actually, i never use Elsa), with the exception of NPCs "received messages" for which i use a specific UK/US voice depending on the type of NPC.

Roger for everything on your other post.

 

[T'kael]

oh! i missed this. I searched the log for "skipping" and overlooked the fact that Carla wasn't mentioned at all...

Never had VoiceAttack, i always use EDDI alone.

ABSOLUTLY sure.
It is the default voice, everything is said with that voice (actually, i never use Elsa), with the exception of NPCs "received messages" for which i use a specific UK/US voice depending on the type of NPC.

Roger for everything on your other post.

The code for fetching voices from the System.Speech synthesizer for 4.0.2 and 4.0.3-b2 are almost identical but there are a few differences in locking statements there as well. I'm not sure if or how those locking statements might produce the difference but I'll try to adjust those to more closely match 4.0.2.

 

[T'kael]

Ok. I've refined the code again and am pleased to announce that EDDI version 4.0.3-b3 is now available. The changelog is available here and the installer is available from the app or from here.

 

[Parduz]

Ok. I've refined the code again and am pleased to announce that EDDI version 4.0.3-b3 is now available. The changelog is available here and the installer is available from the app or from here.

Ok, i've done some "parallel testing" between 4.0.2 and 4.0.3b3, i'll try to write the most precise report.

These are all my installed voices:

What i've done (for each of the two version, with debug log previously enabled) is:

1. Clean the log file with an external editor.
2. Run EDDI and let it "set" for a couple of minutes
3. Pick every listed voice, from the top to bottom, and clicked on the "Test Voice" button
4. Run a script (list below) which try to speak with each voice
5. Close the program
6. Copy the log file elsewhere

Please note that v4.0.3b3 run "inside" Sandboxie with web access enabled, so EDDI can read whatever it wants but what it writes is saved elsewhere (EDDI doesn't know it). I use Sandboxie a lot and AFAIK this shouldn't alter the test.

The "- - TEST ALL VOICES - -" script:

{Voice("Paola."   , "Paola"            )}
{Voice("Roberto." , "Roberto"          )}
{Voice("Fabio."   , "Fabio"            )}
{Voice("Giulia."  , "Giulia"           )}
{Voice("Giorgio." , "IVONA 2 Giorgio"  )}
{Voice("Luca."    , "Luca"             )}
{Voice("Marcello.", "Marcello"         )}
{Voice("Cosimo."  , "Microsoft Cosimo" )}
{Voice("Elsa."    , "Microsoft Elsa"   )}

Passiamo all'inglese.
{Voice("David."   , "Microsoft David"  )}
{Voice("George."  , "Microsoft George" )}
{Voice("Mark."    , "Microsoft Mark"   )}
{Voice("Hazel."   , "Microsoft Hazel"  )}
{Voice("Susan."   , "Microsoft Susan"  )}
{Voice("Zira."    , "Microsoft Zira"   )}

{Voice("Hazel Desktop."   , "Microsoft Hazel Desktop"  )}
{Voice("David Desktop."   , "Microsoft David Desktop"  )}
{Voice("Zira Desktop."    , "Microsoft Zira Desktop"   )}
{Voice("Elsa Desktop."    , "Microsoft Elsa Desktop"   )}

(the "Desktop" versions are there just to see what happens)

Tests report:

NOTE: i completly forgot that Cosimo and Elsa were problematic even in 4.0.2, i asked about them in this post . Please pardon me if this caused confusion .

In 4.0.2 every voice is listed, but Cosimo and Elsa are not working, either by clicking the test button or by running the script.
The script "aborts" when a voice fails, so i had to comment out the two line using Cosimo and Elsa.
All of the "Desktop" voices works too, with the exception of Elsa.

In 4.0.3b3 all the (old) SAPI voices are not listed; all listed voices are working, even Cosimo and Elsa (although i see warnings about them in the log).
With the script you hear that all not listed voices are not found, it says that Zira Desktop is not found (4.0.2 instead speaks with it) and Elsa Desktop fails like in 4.0.2.

Hope this helps, tomorrow i'll try to gather that old laptop for other tests.

 

[nepomuk]

Hello @T'kael, bad news:
The update from v.4.0.3-b2 to v.4.0.3-b3 via EDDI itself does not work for me. It comes the announcement "Update failed! Try again later". All previous updates have always worked without problems.
EDIT:
Ok, it could be that the virus scanner AVAST has a problem with EDDI. Strange, it did not have so far. The update via the .exe file started separately works (with the prompts from AVAST).

 

[Darkcyde]

Ok. I've refined the code again and am pleased to announce that EDDI version 4.0.3-b3 is now available. The changelog is available here and the installer is available from the app or from here.

Sadly, the EDSM issue is still happening for me. This time it got to 10300 of 12630, and it just stopped updating. EDDI is still responsive though. At roughly about 8000 it got really slow, and progressively got slower until it just stopped progressing altogether, even after 5 minutes.

I forgot to turn off verbose logging from the Play() issue I raised, so the log file is massive (~154mb) so even Zipped it won't post here. Still, verbose may be of more help than a normal log. So here's a link to the zip (13mb) on my Dropbox: https://www.dropbox.com/s/796tlyl757ziztx/eddi.zip?dl=0

Edit: Hmm, I just noticed when turning off verbose logging, that the EDDI main tab says that verbose logging will create a Zip on my desktop. However, that doesn't happen, I had to Zip it myself. Was that functionality removed? If so, the main tab text needs updating. If not, then that may be another bug.

 

[T'kael]

Sadly, the EDSM issue is still happening for me. This time it got to 10300 of 12630, and it just stopped updating. EDDI is still responsive though. At roughly about 8000 it got really slow, and progressively got slower until it just stopped progressing altogether, even after 5 minutes.

I forgot to turn off verbose logging from the Play() issue I raised, so the log file is massive (~154mb) so even Zipped it won't post here. Still, verbose may be of more help than a normal log. So here's a link to the zip (13mb) on my Dropbox: https://www.dropbox.com/s/796tlyl757ziztx/eddi.zip?dl=0

Edit: Hmm, I just noticed when turning off verbose logging, that the EDDI main tab says that verbose logging will create a Zip on my desktop. However, that doesn't happen, I had to Zip it myself. Was that functionality removed? If so, the main tab text needs updating. If not, then that may be another bug.

It's actually a different EDSM issue. The problem is that we went too fast and ran into EDSM's rate limit. Hitting EDSM's rate limit is is a very bad thing, unfortunately, and the consequence is an hour-long cool-down before EDDI can make another request to that endpoint. On the plus side though, your log did reveal that EDDI was making more requests to EDSM than intended. I'll work on that.

 

[T'kael]

Ok, i've done some "parallel testing" between 4.0.2 and 4.0.3b3, i'll try to write the most precise report.

These are all my installed voices:
View attachment 358288
What i've done (for each of the two version, with debug log previously enabled) is:

1. Clean the log file with an external editor.
2. Run EDDI and let it "set" for a couple of minutes
3. Pick every listed voice, from the top to bottom, and clicked on the "Test Voice" button
4. Run a script (list below) which try to speak with each voice
5. Close the program
6. Copy the log file elsewhere

Please note that v4.0.3b3 run "inside" Sandboxie with web access enabled, so EDDI can read whatever it wants but what it writes is saved elsewhere (EDDI doesn't know it). I use Sandboxie a lot and AFAIK this shouldn't alter the test.

The "- - TEST ALL VOICES - -" script:

{Voice("Paola."   , "Paola"            )}
{Voice("Roberto." , "Roberto"          )}
{Voice("Fabio."   , "Fabio"            )}
{Voice("Giulia."  , "Giulia"           )}
{Voice("Giorgio." , "IVONA 2 Giorgio"  )}
{Voice("Luca."    , "Luca"             )}
{Voice("Marcello.", "Marcello"         )}
{Voice("Cosimo."  , "Microsoft Cosimo" )}
{Voice("Elsa."    , "Microsoft Elsa"   )}

Passiamo all'inglese.
{Voice("David."   , "Microsoft David"  )}
{Voice("George."  , "Microsoft George" )}
{Voice("Mark."    , "Microsoft Mark"   )}
{Voice("Hazel."   , "Microsoft Hazel"  )}
{Voice("Susan."   , "Microsoft Susan"  )}
{Voice("Zira."    , "Microsoft Zira"   )}

{Voice("Hazel Desktop."   , "Microsoft Hazel Desktop"  )}
{Voice("David Desktop."   , "Microsoft David Desktop"  )}
{Voice("Zira Desktop."    , "Microsoft Zira Desktop"   )}
{Voice("Elsa Desktop."    , "Microsoft Elsa Desktop"   )}

(the "Desktop" versions are there just to see what happens)

Tests report:

NOTE: i completly forgot that Cosimo and Elsa were problematic even in 4.0.2, i asked about them in this post . Please pardon me if this caused confusion .

In 4.0.2 every voice is listed, but Cosimo and Elsa are not working, either by clicking the test button or by running the script.
The script "aborts" when a voice fails, so i had to comment out the two line using Cosimo and Elsa.
All of the "Desktop" voices works too, with the exception of Elsa.

View attachment 358286

In 4.0.3b3 all the (old) SAPI voices are not listed; all listed voices are working, even Cosimo and Elsa (although i see warnings about them in the log).
With the script you hear that all not listed voices are not found, it says that Zira Desktop is not found (4.0.2 instead speaks with it) and Elsa Desktop fails like in 4.0.2.
View attachment 358291

Hope this helps, tomorrow i'll try to gather that old laptop for other tests.

It will take me some time to study this and your prior posts. Thank you. If you can find the time then it would be helpful to have a table clearly summarizing the results from your powershell script tests as well (which voices were found and which voices were rendered successfully with which synthesizer)

 

[Darkcyde]

I've been using "DB Browser for SQLite" to repair my EDDI database today. I've completed one system OK. I then decided to look for all systems that have a null entry for the system address, with the intent to use EDSM to correct those. However, I've found that I have almost 1900 systems where "systemAddress":null exists. I think that's just a few too many for me to bother with, even if it does help with obtaining my EDSM logs.

 

[Neilski]

However, I've found that I have almost 1900 systems where "systemAddress":null exists. I think that's just a few too many for me to bother with, even if it does help with obtaining my EDSM logs

Yeah 1900 is a non-starter for a human to fix, but just extra run time if you can script the fix.

 

[Parduz]

I think that's just a few too many for me to bother with

It feels a bit awkward to ask this to you but .... do you need an hand in scripting what you need to do?

If you can find the time then it would be helpful to have a table clearly summarizing the results from your powershell script tests as well (which voices were found and which voices were rendered successfully with which synthesizer)

Ok, i'll do it.

 

[Darkcyde]

Yeah 1900 is a non-starter for a human to fix, but just extra run time if you can script the fix.

I just realised that Fleet Carriers are recorded in the system data, and they have a system address of null. So I guess a great many of that 1900 are probably just systems with a carrier in.

 

[Darkcyde]

It feels a bit awkward to ask this to you but .... do you need an hand in scripting what you need to do?

That would be appreciated, as I don't know much about SQL scripting (ie, none at all). I don't know how easy/hard it may be though, as I've just found that some entries of "systemAddress" in the system data actually need to be null, for Fleet Carriers.

 

[Saltyseadog]

Hello @T'kael, bad news:
The update from v.4.0.3-b2 to v.4.0.3-b3 via EDDI itself does not work for me. It comes the announcement "Update failed! Try again later". All previous updates have always worked without problems.
EDIT:
Ok, it could be that the virus scanner AVAST has a problem with EDDI. Strange, it did not have so far. The update via the .exe file started separately works (with the prompts from AVAST).

I just tried to update from b2 to b3 and got the same "update failed" message. On a repeat try I got a Windows pop up saying that update b3 was incompatible with Windows 10 64 bit and to ask if a 64bit version was available.

 

[nepomuk]

I just tried to update from b2 to b3 and got the same "update failed" message. On a repeat try I got a Windows pop up saying that update b3 was incompatible with Windows 10 64 bit and to ask if a 64bit version was available.

This statement can also be found in my first log. With the update via the .exe file everything is OK in the log.