[Bug] Voice issues and more docuementation needed for migration to new voice system

[FingerlessGlov3s]

Summary

We've added voice_internal and setr sv_mumble true to the server.cfg but the old pma-voice resource which uses mumble doesn't seem to work, I assumed setr sv_mumble true was going to make all the mumble natives work. Not sure if others are having issues with voice not working, we're currently using pma-voice. Nothing seems to be erroring, but voice isn't really working. I have a suspicion it's related to playerids getting reused and mumble isn't liking it maybe.

I've gone reading https://docs.fivem.net/docs/scripting-manual/voice/#fivem-for-gtav-enhanced to convert over to the new voice system as it's deprecated, is there natives missing or is it not quite 1:1 as it were yet.

These are my findings about trying to migrate, it could be there are new natives for this but not mentioned on the scripting-manual at this time. I could be wrong with my findings, or misunderstand of how it supposed to work, it might be useful if there was a cfx native reference that has only enhanced natives within it, to know what's not available or getting deprecated.

A. Missing natives - feature is lost, no replacement

#	Deprecated native	What it provides	What is needed
A1	MumbleSetAudioInputDistance	Per-speaker talk range (whisper / normal / shout)	maxDistance is fixed per channel at creation, so range is a property of the channel, not the speaker. The only workaround is one channel per range tier with every player a member of all tiers and unmuted in one - which multiplies spatial indexes by the tier count. Wanted: a per-member range, e.g. SetPlayerVoiceRangeInChannel(channelId, clientId, distance), or a distance argument on add
A2	MumbleSetVolumeOverrideByServerId	Per-listener, per-speaker volume	Used for radio/call volume sliders, per-channel volume when monitoring multiple channels, ducking a lower-priority speaker during a priority transmission, and letting a player locally quieten one individual. The new API has no volume control of any kind
A3	MumbleSetSubmixForServerId	Attaching a voice stream to an audio submix	The AUDIO-namespace submix natives (CreateAudioSubmix, SetAudioSubmixEffectRadioFx, AddAudioSubmixOutput) still exist, but nothing routes a voice stream into one. There is currently no way to apply any audio effect to voice - this removes radio filter effects and megaphone/PA effects entirely
A4	MumbleIsPlayerTalking	Talking state	No documented way to learn who is currently speaking. Drives talking indicators, HUD mic activity, push-to-talk edge detection, and recording/bodycam resources
A5	MumbleSetAudioInputIntent	Speech vs music encoding hint	No replacement

---

B. Undocumented behaviour - blocks correct implementation

#	Question	Why it matters
B1	If a speaker is unmuted in two channels that a listener is a member of, does the listener receive one stream or two?	The standard radio requirement is: a nearby listener on the same radio channel hears the speaker once, through the radio, with the radio effect - while a nearby listener not on that channel hears them normally. A transmitting player is necessarily unmuted in both their proximity channel and their radio channel. If delivery is not deduplicated, the listener hears a dry copy and an effected copy simultaneously, and there is no per-channel volume to suppress either
B2	Mode 0 (Non-spatial) states 2D output positioning "must be controlled with the client voice API"	That client voice API is not documented anywhere
B3	Is there a per-channel member cap, and what is the expected cost at scale?	The A1 workaround puts every player into every range tier. On a high-population server (500–1000 concurrent) that is several thousand memberships. Whether that is viable is unknown, and there is no guidance
B4	Do channels survive a resource restart, and how are orphans reclaimed?	Channels appear to be owned by the server process, so a resource restart loses the script's ID cache while the channels persist. There is no native to enumerate existing channels, so orphaned channels cannot be found or cleaned up, and repeated restarts walk toward the 65535 cap

Issue type

Both / Unsure

Repro rate

Always

Server build version

FXServer-early-access b96 win32

OS

No response

CPU

No response

GPU

No response

RAM

No response

Storage type

None

Connection type

None

ISP and bandwidth

No response

DxDiag

No response

Network graph

No response

Platform

None

OS version / distribution

No response

CPU

No response

RAM

No response

Using txAdmin?

None

Hosting provider

None

Machine type

None

/perf endpoint output

No response

DDoS protection

No response

ulimit -n value (Linux only)

No response

Docker Compose file (Docker only)

No response

Steps to Reproduce

1. use pma-voice on Enhanced or look at the new native docs to try workout how to convert it to the new voice system

Expected Behavior

pma-voice to work until all the new voice system contains all feature set seen in the old mumble system.

Actual Behavior

Mumble voice isn't working correctly at least in our server.

Evidence

No response

Additional Context

Would be nice to hear from other server owners to know if their also having issues with voice, and if they are using pma-voice

[FingerlessGlov3s]

Little update using mumble legacy natives: Lots of players are talking now, we're not sure why other than we think it might be related to turning off Echo Cancellation, Mic Level Normalization and Noise Suppression, seems to be the common denominator, for those it is working for now.

The 3d voice is working, but the toggling of the range ain't working but it's better than nothing!

[Trebor96]

Still an issue! Voice is very hit and miss, works for a limited amount of people per restart of a server...

[wolfram-cfx]

Hi, we are preparing more fixes related to the mumble voice api.
I'll answer the questions in the behavior they have after most issues in the compatibility layer are fixed.

Do channels survive a resource restart, and how are orphans reclaimed?

We have temporary channels, these are mainly used by the mumble compatibility layer, but can be used by the scripts as well. These type of channel deletes itself when the last player leaves.

Is there a per-channel member cap, and what is the expected cost at scale?

There is no limit and there is no expected cost at scale.

Mode 0 (Non-spatial) states 2D output positioning "must be controlled with the client voice API"

We'll open the voice client api once we are happy with the stability of the current voice output to rule out conflicts.

If a speaker is unmuted in two channels that a listener is a member of, does the listener receive one stream or two?

The rule is that no voice packet is every duplicated. It is always one sender and then n receivers and no receiver will receive the same packet twice. The job of possible transformations for this audio data will be done by the client voice api.

[wolfram-cfx]

To start a discussion about the missing natives:

MumbleSetAudioInputDistance

Since channels now are spatial as well this overlaps with the responsibility of a channel. So not sure if both ways should be done.

MumbleSetVolumeOverrideByServerId

Needs to be handled by the client voice api on the receiver end. We will need to think about some meta info to indicate what kind of voice data the player received.

MumbleIsPlayerTalking

This has a game native that does the same thing (NETWORK_IS_PLAYER_TALKING). But I agree there probably could be a custom one as well.

MumbleSetAudioInputIntent

Should also be handled by the client voice api.