Repository navigation
3D Directional Audio #44
Description
This is probably a long shot, but I saw this video about an early Retail version of SS:TSE (patch 1.50 beta) : https://www.youtube.com/watch?v=TQI8nNh8Azg
It made me wonder: Could 3D audio be restored for the Steam release of the Serious Sam The Second Encounter?
Activity
LethalManBoob commentedon Feb 26, 2026
openal soft works for me with dsoal
vlad54rus commentedon Feb 28, 2026
It works but limited to 2D only. No 3D sound positions for proper HRTF. No multichannel output support.
HyperVirusProfi commentedon Apr 7, 2026
Так же с DSOAL пытался крутиться, она у меня работала на более старых версиях патча с каким-то модом вроде, перетаскивал в классику чисто - на некоторых локациях просто космос, в других косяк
\да было бы круто такое юзать, но хз как это для сэма реализовать
adhdandy commentedon Oct 12, 2026
Findings on how the 1.50 beta differs from the Steam/1.10 engine (from the public 1.10 source, the beta's Engine.dll exports, and DSOAL debug logs of both builds):
1.10 source / Steam build
CSoundObject::Update3DEffects()→ per-ear volume, filter, delay, pitch, phase).DSBCAPS_CTRL3Dhits, listener at (0,0,0), exactly two bufferSetPositioncalls at (-1,0,0) and (+1,0,0). Nothing per-source. This is why DSOAL only gives 2D results here.1.50 beta
Update3DEffects, a voice pool (CommitPlayingBuffer,ReleasePlayingBuffer, pending list),SetEnvironment(SndEnvFX), and cvars likesnd_iMaxVoices,snd_bUseHeadphones,snd_iEmulationType.DSBCAPS_CTRL3Dbuffers; per-bufferSetPositionwith continuously changing, roughly unit-length vectors;SetMode(head-relative on some buffers, disabled on others);SetVelocity;SetDopplerFactor;SetRolloffFactor(0). Directional audio is clearly audible on headphones.Conclusion
The beta did per-source positional audio through DirectSound3D. It passes the engine's own listener-relative direction per sound in head-relative mode, does distance attenuation itself (rolloff 0), and leaves panning to the API. The Steam build doesn't. Restoring this means feeding per-sound direction into a 3D-capable backend.
Not checked: how the beta's
snd_iEmulationType/snd_bUseHeadphonesvalues map to behavior, and the exactSetModesemantics per sound type.DreamyCecil commentedon Oct 12, 2026
I've known about the differences in the sound API between 1.07/1.10 and 1.50 from my own reverse engineering and from looking at leaked Last Chaos source code for reference. You can look at its sound API yourself, I'm pretty sure it's basically identical to clean Serious Engine 1.50: https://github.com/openlastchaos/lastchaos-source-client/tree/master/Engine/Sound
Other than that, it would be significantly difficult to implement proper 3D audio when the engine itself does not support it. It would need an entire new sound interface on top of the game that would hijack any attempts at playing the sounds and redirecting them to the new API. But I'm not sure how hard it would be to keep track of each individual sound object to keep them in sync with sound objects in the new API in this case.
adhdandy commentedon Oct 14, 2026
I've spent some time investigating the differences in 3D audio between the TSE 1.50 beta and the Steam release, and I've documented my findings and a possible implementation approach here:
Research: https://github.com/adhdandy/serious-sam-ds3d-findings
I've also documented a potential approach for restoring per-sound spatialization through runtime patches and shadow buffers, without replacing the stock Engine.dll. There are still open questions around sound-object synchronization and playback lifecycle handling, and the proposed implementation has not yet been tested in-game.
For transparency, this is my first reverse-engineering project, and I relied heavily on AI tools throughout the investigation, including for disassembly analysis, interpreting engine behavior, and developing the proposed design. I've tried to distinguish observed evidence from hypotheses, but the findings and design should be considered a starting point for further verification rather than a proven implementation.
I'd be very interested in your thoughts, particularly on whether the proposed approach to synchronizing the shadow buffers with the engine's sound objects seems viable. If this looks promising, I'd be happy to continue investigating the outstanding questions.
The two most useful starting points are findings/99-summary-and-open-questions.md and findings/08-design.md.
DreamyCecil commentedon Oct 14, 2026
I'm going to be honest, this is way too much technical information that I don't understand in the slightest. And I don't really understand the point in this research when new code would have to be written on top of the game either way.
The way I see it right now - if I wanted to implement proper 3D audio to the patch, I would begin by implementing it in the engine itself with a fork of 1.10 and then just copying that code to the patch and patching a few functions to interact with my own sound system instead of the vanilla one from the engine. But I don't know how to do that yet because I've never worked with proper 3D audio before.
LethalManBoob commentedon Oct 14, 2026
I always figured you can just pass the audio off to openal and let that do the processing for hrtf. Since dsoal on og serious sam does just that.
adhdandy commentedon Oct 14, 2026
Thanks for the honest reply, and sorry for the wall of text. Short version of what's useful for your 1.10 fork plan:
How the 1.50 beta does 3D audio (from its Engine.dll plus a DSOAL call log, not yet tested in-game on Steam):
CommitDeferredSettingsper frame.findings/99-summary-and-open-questions.md, section 1.The sync question: in the beta, a sound's playback state lives in the DS buffer. Each frame it asks the buffer whether it is still playing, and when a non-looping sound finishes, the engine object gets stopped. The software mixer simply skips these sounds.
One trap in Steam/1.10: you can't mute a sound by setting its volume to 0. The mixer stops non-looping zero-volume sounds after one block. You have to skip the sound in
MixSoundinstead (SoundMixer.cpp:624-659).If you use the Last Chaos 1.50-era sound source, checking my numbers against it would be the quickest way to confirm or break my findings.
adhdandy commentedon Oct 18, 2026
Quick follow up, I started my implementation of spatial audio here: https://github.com/adhdandy/SuperProject/tree/ds3d-spatial
Currently unfinished, but in testing it seems to work fine (so far). I will post here when I feel it's ready.
adhdandy commentedon Oct 19, 2026
Update: the spatial audio mode on the
ds3d-spatialbranch is now ready for testing.It restores per-sound directional audio for the Steam release of TSE (1.07), like the 1.50 beta. The stock
Engine.dllis not replaced. With the option off, the game behaves as vanilla.Requirements (all of these are needed):
Binfolder. It bundles OpenAL Soft and turns DirectSound3D positions into HRTF/binaural output.snd_iInterface 2). Plain DirectSound has no 3D listener, so the mode stays off there.snd_bDS3DSpatial 1in the console. It is off by default, so nothing changes unless you set it. It's saved with your settings.Use headphones. The effect is barely noticeable on speakers. Try a level with lots of Beheaded Kamikaze, i.e. "The City of the Gods".
Install: s3d-spatial-beta. Use the Core and exe files from the release together. A mix of different builds will not load.
What to expect:
snd_iEmulationType(0/1/2) switches between the default algorithm and the two HRTF modes. It applies live.Limits:
What I'd like to know:
DSOAL_LOGFILEandDSOAL_LOGLEVELenvironment variables) and say what you did.jyetcv commentedon Oct 19, 2026
@adhdandy
I'm using dsoal with Proton and it sounds right to me.