OnCourse Software

I've excluded all robot voices in PF3 and noticed that it is saved in VoiceSetExclusions.dat file.
I attached my version here, just unzip it and copy it to ...\PF3\Data\

Cheers!

I agree with many of your exclusions, but some of them sound really nice (e.g. DC) so it would be a loss to exclude them.
As indicated (but explicitly stated) in the other thread, I would recommend listening to longer examples, e.g. pfchat144*.wav

ThomasAH wrote:

I agree with many of your exclusions, but some of them sound really nice (e.g. DC) so it would be a loss to exclude them.

I checked all voice files, here is my rating of the voices you excluded:

Code: Select all AC low AD good AE low AF low CA low CB low CC low CD low CF medium CG low DC good DD good DE good EA low EB medium EC good ED medium EE low EF low EG medium ER medium ES low ET low EU low EV medium EW low EX medium EY low MA low MB low MC low MD medium ME good MF low MG medium PA good WA low WB medium WC low WD low WE low WF low WG low WH medium WI medium WQ low WR medium WS medium WT medium WU medium WV good WW low WX medium WY good WZ good XC low XD medium ZA low ZB medium ZC low

I attacked a zip with two files:
VoiceSetExclusions/just-good/VoiceSetExclusions.dat excludes 50 voices (with my personal rating of low or medium)
VoiceSetExclusions/with-medium/VoiceSetExclusions.dat excludes 31 voices (with my personal rating of low) For comparison: You excluded 60 voices, I consider 10 of them as good.

All others generally sound "good" for me, but for my own setup I excluded KC for his exaggerated pronunciation of 3 as Tree. Even if technically correct I just don't like to hear it that strong (and that often, as he is very active in the regions I fly).

What I noticed when listening to the voices is, that many female voices do not sound good. Somehow like recordings of male speakers, already downsampled to 8bit/11025Hz have been altered to sound female, and thus adding sound artifacts.

Thank you both for sharing your lists of excluded voice sets.

When I look at them it proves Dave March was right when he said (in some other thread) the rating of the voice sets is highly subjective: though I also dislike some of the voices I would never exclude so many of them!

So I guess everyone should decide for himself which voice sets to exclude. But just listening to one wav-file is not enough for a profound decision.

Here is how it can be done:

Go to an airport in the USA (because there are the most voice sets).
Set AI traffic to zero in the sim (to concentrate on your pilot's voice)
Plan a flight starting from that airport.
Set your Comm to Clearance or Ground.
In PF3:
Set CP Mode 2.
Enter a long SID name including letters and numbers.
Choose the voice you want to test as your pilot's voice.
Choose the flight plan and connect to FS.
Now with a few hotkey presses you can easily go through the initial clearance, the pushback clearance and the taxi clearance. That should be enough to decide if you like the voice or not. Incidentally you might already be able to rate the ATC voices talking to you.
Disconnect PF3.
Choose the next voice as your pilot's voice.
Again choose the flightplan and connect to FS.
Test the next voice (And maybe incidentally another ATC voice - the more voice sets per region the higher the chance of a different one each time. That's why you should start in the USA. You might want to move to the UK later on).

This takes about 3 minutes per voice but gives a better result than just listening to one wav-file.

Regards

Ralf

RALF9636 wrote:

This takes about 3 minutes per voice but gives a better result than just listening to one wav-file.

3 minutes per voice would be 6 hours ...
For my checks I listened to pfchat144 and pfchat235 (though many recordings did not include the long part about contacting AETI on www.simpilot.com)

An alternative (and probably recommended, if you do not want to have false positives) way would be to note the voice ids when you notice that you do not like them during a flight, but that would only gradually improve the personal experience.

ThomasAH wrote:

RALF9636 wrote:

This takes about 3 minutes per voice but gives a better result than just listening to one wav-file.

3 minutes per voice would be 6 hours ....

With the ATC voices you get incidentally you might make it in 4-5 hours...

ThomasAH wrote:

I agree with many of your exclusions, but some of them sound really nice (e.g. DC) so it would be a loss to exclude them.
As indicated (but explicitly stated) in the other thread, I would recommend listening to longer examples, e.g. pfchat144*.wav

Wow, I've just realised that I've excluded three good Australian voices, and left robot ones. DA and DB are robot voices.

I think I'm the one who first called them Robot Voices! Thank you so much for addressing this. I understand the appeal of using synthesized voices...you don't need to hire actors or go to a recording studio...you can just make them from your station...BUT...they sound just awful if you will forgive me...some are painful to hear...

The live recording voices on the other hand are stellar...The difference is stark, and I can't imagine users preferring the robots...

To make things simpler, determine which voices were created by computer synthesis rather than by live recording...and just delete those...

Thanks again for taking the time to get them out...I see it is a work in progress and I'm excited about using the exclude file when it's sorted.

mllgrennman wrote:

I think I'm the one who first called them Robot Voices! Thank you so much for addressing this. I understand the appeal of using synthesized voices...you don't need to hire actors or go to a recording studio...you can just make them from your station...BUT...they sound just awful if you will forgive me...some are painful to hear...

You don't need voice actors for good sounds. I started a project for ProATC/X to make new voices in all accents. The community has lots of people willing to speak-in a set. And it's cool to have your own voice in-sim as the pilot!

Yes I agree...but recorded voices are still understandably more realistic than synthesized.