docs: added troubleshooting section about Geforce Now

This commit is contained in:
alterNERDtive 2022-08-23 22:26:11 +02:00
parent 1357bde807
commit 3775d3f911
Signed by: alterNERDtive
GPG key ID: 547787A4FE6533F1

View file

@ -1,91 +1,100 @@
# Troubleshooting
This will fill up gradually with Troubleshooting tips as people run into common
ones.
## VoiceAttack does not understand me / mishears me / fires random commands
There is [a thread on the VoiceAttack
forums](https://forum.voiceattack.com/smf/index.php?topic=2667.msg12197#msg12197)
on how to set up your microphone and the speech recognition engine to work best.
If your microphone is bad and you still get erroneous recognitions when you are
not speaking it is probably going to recognize the same command every time. You
can remedy that by blocking the voice trigger. One-syllable triggers are
especially prone to misrecognition.
1. Create a new command in your custom profile.
1. Set the “when I say” field to the trigger that gets misrecognized.
Adding the “Other” → “VoiceAttack Action” → “Ignore an Unrecognized Word or
Phrase” action will also hide it from the VoiceAttack log when it is (wrongly)
recognized. You might or might not want that.
Example for the “cruise” voice trigger of the Supercruise command:
![[troubleshooting-remove-trigger.png]]
Alternatively you can raise the minimum confidence level and call the underlying
command to make misfires less likely:
![[troubleshooting-raise-min-confidence.png]]
There are a few examples in the [Custom Profile
Example](../installing#use-the-profile-example).
## VoiceAttack recognizes a command, but doesnt do anything in game
Make sure you have a keyboard bind for whatever the command is supposed to do as
outlined in [[Installing#Set Elite Keyboard Binds]].
## VoiceAttack talks over the COVAS voice
There is no way to know for sure when the ingame COVAS is talking to you, so
there is no way to always prevent the two from speaking over each other.
You can however either disable TTS responses for events that you know will clash
(or [file a feature
request](https://github.com/alterNERDtive/VoiceAttack-profiles/issues/) if one
is not optional yet). Alternatively, if you prefer the info given by VoiceAttack
over the ingame COVAS, you can deactivate its response to these events in the
ingame Audo settings:
![[Elite-COVAS.png]]
While youre in there you might as well get rid of the spoken FSD countdown that
# Troubleshooting
This will fill up gradually with Troubleshooting tips as people run into common
ones.
## VoiceAttack does not understand me / mishears me / fires random commands
There is [a thread on the VoiceAttack
forums](https://forum.voiceattack.com/smf/index.php?topic=2667.msg12197#msg12197)
on how to set up your microphone and the speech recognition engine to work best.
If your microphone is bad and you still get erroneous recognitions when you are
not speaking it is probably going to recognize the same command every time. You
can remedy that by blocking the voice trigger. One-syllable triggers are
especially prone to misrecognition.
1. Create a new command in your custom profile.
1. Set the “when I say” field to the trigger that gets misrecognized.
Adding the “Other” → “VoiceAttack Action” → “Ignore an Unrecognized Word or
Phrase” action will also hide it from the VoiceAttack log when it is (wrongly)
recognized. You might or might not want that.
Example for the “cruise” voice trigger of the Supercruise command:
![[troubleshooting-remove-trigger.png]]
Alternatively you can raise the minimum confidence level and call the underlying
command to make misfires less likely:
![[troubleshooting-raise-min-confidence.png]]
There are a few examples in the [Custom Profile
Example](../installing#use-the-profile-example).
## VoiceAttack recognizes a command, but doesnt do anything in game
Make sure you have a keyboard bind for whatever the command is supposed to do as
outlined in [[Installing#Set Elite Keyboard Binds]].
## VoiceAttack talks over the COVAS voice
There is no way to know for sure when the ingame COVAS is talking to you, so
there is no way to always prevent the two from speaking over each other.
You can however either disable TTS responses for events that you know will clash
(or [file a feature
request](https://github.com/alterNERDtive/VoiceAttack-profiles/issues/) if one
is not optional yet). Alternatively, if you prefer the info given by VoiceAttack
over the ingame COVAS, you can deactivate its response to these events in the
ingame Audo settings:
![[Elite-COVAS.png]]
While youre in there you might as well get rid of the spoken FSD countdown that
is off by one second …
## This doesnt work (well) with my HCS pack
My profiles are designed from the ground up to work with whatever else you are
doing with VoiceAttack; that is the reason for importing them into your own
My profiles are designed from the ground up to work with whatever else you are
doing with VoiceAttack; that is the reason for importing them into your own
custom profile instead of selecting e.g. `EliteAttack` as your active profile.
HCS on the contrary explicitly expects you do exclusively use HCS with
VoiceAttack. There is the rudimentary way of including simple profiles into
theirs, but the mechanism falls flat in many places. For example you cannot tell
HCS on the contrary explicitly expects you do exclusively use HCS with
VoiceAttack. There is the rudimentary way of including simple profiles into
theirs, but the mechanism falls flat in many places. For example you cannot tell
HCS to run an included profiles startup command.
So, in order to mostly make stuff work, you need to treat the HCS profile as
So, in order to mostly make stuff work, you need to treat the HCS profile as
your “custom” profile as per this documentation.
1. Include `alterNERDtive-base` and all profiles you want to use into the active
1. Include `alterNERDtive-base` and all profiles you want to use into the active
HCS profile.
1. Include a custom profile that has a startup command with a voice trigger of
1. Include a custom profile that has a startup command with a voice trigger of
your choosing, e.g. “load included profiles”.
1. Set up said startup command as you would normally.
1. Every time you start VA or change profiles, you will have to manually say
1. Every time you start VA or change profiles, you will have to manually say
“load included profiles”.
That will make most things work. Conflicts may arise if HCS happens to have
voice triggers that are the same as mine, in which case their command will take
That will make most things work. Conflicts may arise if HCS happens to have
voice triggers that are the same as mine, in which case their command will take
priority.
**Note on TTS**: EDDIs TTS (used by my profiles) and HCS TTS / recorded voice
lines act 100% independently. That means they will frequently “speak over each
**Note on TTS**: EDDIs TTS (used by my profiles) and HCS TTS / recorded voice
lines act 100% independently. That means they will frequently “speak over each
other”. There is no way to alleviate this.
EDDI does have a mechanism to detect if it is currently speaking it sets a
corresponding VoiceAttack variable. HCS neither does anything similar nor checks
if EDDI is speaking to prevent conflicts. Refer to them if you want that
EDDI does have a mechanism to detect if it is currently speaking it sets a
corresponding VoiceAttack variable. HCS neither does anything similar nor checks
if EDDI is speaking to prevent conflicts. Refer to them if you want that
changed.
## This does not work with Geforce Now
Nope. Just wont. Geforce Now obviously has Elites files, journals and keybinds
stored on some random PC in the cloud. Your local VoiceAttack has no way of
accessing those.
If you can get VoiceAttack to run _on Geforce Now_ in parallel to Elite, I dont
see why it wouldnt work.