Acoustics: per-source room estimate (sources in another room get their own reverb and send level) #857

New issue

Closed

opened 2026-10-19 01:56:41 +00:00 by sky_line · 1 comment

sky_line commented

2026-10-19 01:56:41 +00:00

Owner

Follow-up to #683 (RT room acoustics).

Today: the room estimate comes only from rays around the listener, and every source sends to one shared reverb. A source in the next room, or outside, sounds like it's in your room; the send occlusion added recently only darkens and attenuates it.

Proposal:

Per-source probe: for each audible in-world source (nearest / loudest N), cast a small set of rays around the source (e.g. 8-16, rotated per frame, smoothed over ~1 s) and estimate its local enclosure, room size and absorption with the same model as the listener's.
Same room (similar estimate and the direct path is clear or only lightly occluded): the source uses the listener's reverb, as today.
Different room: the source sends to a second reverb slot configured from its own estimate (OpenAL Soft allows 4-6 aux sends / slots). The "other room" reverb is heard through the occluded path. Coupled spaces: a hall heard through a doorway sounds like a hall.
Send levels per source come from its own enclosure, so a source outdoors adds little reverb even when you're indoors.

Constraints:

Ray budget a few dozen extra rays per frame at most, sharing the existing dispatch and readback.
Slots are limited, so cluster sources by room estimate.
Off when acoustics is off; stands down with RT like the rest of the feature.

Done when: a sound in an adjacent hall or outdoors is clearly distinguishable from one in your own room in a blind listen, at under ~0.05 ms extra GPU on an RTX 3060-class card.

Follow-up to #683 (RT room acoustics). **Today:** the room estimate comes only from rays around the listener, and every source sends to one shared reverb. A source in the next room, or outside, sounds like it's in *your* room; the send occlusion added recently only darkens and attenuates it. **Proposal:** - **Per-source probe:** for each audible in-world source (nearest / loudest N), cast a small set of rays around the source (e.g. 8-16, rotated per frame, smoothed over ~1 s) and estimate its local enclosure, room size and absorption with the same model as the listener's. - **Same room** (similar estimate and the direct path is clear or only lightly occluded): the source uses the listener's reverb, as today. - **Different room:** the source sends to a **second reverb slot** configured from its own estimate (OpenAL Soft allows 4-6 aux sends / slots). The "other room" reverb is heard through the occluded path. Coupled spaces: a hall heard through a doorway sounds like a hall. - **Send levels per source** come from its own enclosure, so a source outdoors adds little reverb even when you're indoors. **Constraints:** - Ray budget a few dozen extra rays per frame at most, sharing the existing dispatch and readback. - Slots are limited, so cluster sources by room estimate. - Off when acoustics is off; stands down with RT like the rest of the feature. **Done when:** a sound in an adjacent hall or outdoors is clearly distinguishable from one in your own room in a blind listen, at under ~0.05 ms extra GPU on an RTX 3060-class card.

sky_line added the

feature

status

up-next

labels

2026-10-19 01:56:41 +00:00

sky_line added this to the Slipstream project

2026-10-19 01:56:41 +00:00

sky_line referenced this issue

2026-10-19 01:56:46 +00:00

Explore: Steam Audio (ray-traced occlusion, reverb, pathing, HRTF) for in-world sounds #683

sky_line referenced this issue