Audio Component Throws Error When System Audio Is Disabled

[fsj2009yx]

My mods and Everest install are up to date

Yes

I have recreated the bug with only Everest OR a minimum number of mods enabled

Yes

Describe the bug

When the system audio device is disabled or unavailable, the game fails during startup while initializing FMOD.

The exception is thrown in:

system.initialize(...)

with the following error:

FMOD Failed: ERR_OUTPUT_INIT

As a result, the entire game crashes instead of falling back to a silent / no-audio mode.

(I think there should have a catch intead of panicing the whole process)

More interestingly, the game only works when I connect a Bluetooth headset, and the built-in laptop audio device also triggers the same error.

Steps to reproduce

the Game init Crash come with logs:

Ver 1.4.0.0-fna [Everest: 6305-azure-1b9d1-stable]
05/22/2026 17:08:45
System.Exception: FMOD Failed: ERR_OUTPUT_INIT (Error initializing output device.)
   at Celeste.Audio.CheckFmod(RESULT result) in Celeste.Mod.mm/Patches/Audio.cs:line 42
   at Celeste.Audio.Init() in Celeste.Mod.mm/Patches/Audio.cs:line 73
   at Celeste.GameLoader.LoadThread() in Celeste.Mod.mm/Patches/GameLoader.cs:line 87
   at Celeste.RunThread.RunThreadWithLogging(Action method) in Celeste.Mod.mm/Patches/RunThread.cs:line 39

Expected behavior

I Think the Game should fallback to no-audio mode ,even if it cannot automatically choose audio output device when Game is running

Operating System

Windows11 25H2

Everest Version

6305

Mods required to reproduce

No response

Additional context

No response

[SnipUndercover]

This crash is there in vanilla and has been there since day one to my knowledge.
I don't think Celeste was designed to work with an uninitialized sound system. We do have patches to completely stub the FMOD subsystem out when running in headless mode (see Celeste.Mod.mm/Patches/Audio.cs) but this very much feels like a hacky workaround, especially since the entire FMOD API is still available for mods. I don't know whether FMOD handles null pointers gracefully though.

I personally don't consider this a bug, but the rest of the team may disagree.