# Water ambience sound sources

These loops use field recordings released under
[CC0 1.0](https://creativecommons.org/publicdomain/zero/1.0/) (public domain).
Each source page's CC0 link was checked when its public high-quality MP3
preview was downloaded on September 30, 2026. The previews, rather than the
original uploaded WAV/MP3 files, are the source audio used here.

| File | Recording | Author | Source excerpt | Public HQ preview |
| --- | --- | --- | --- | --- |
| `river.ogg` | [Stream.wav](https://freesound.org/people/sterferny/sounds/382322/) | sterferny | 4.0–22.0 s | [382322_3899984-hq.mp3](https://cdn.freesound.org/previews/382/382322_3899984-hq.mp3) |
| `lake.ogg` | [Lake waves.wav](https://freesound.org/people/Anya_Media/sounds/523392/) | Anya_Media | 18.0–36.0 s | [523392_2010973-hq.mp3](https://cdn.freesound.org/previews/523/523392_2010973-hq.mp3) |
| `waterfall.ogg` | [NIAGARA WATERS ROARING.mp3](https://freesound.org/people/KevinT1001/sounds/379905/) | KevinT1001 | 3.0–21.0 s | [379905_571017-hq.mp3](https://cdn.freesound.org/previews/379/379905_571017-hq.mp3) |

[`scripts/build-water-audio.py`](../../../scripts/build-water-audio.py) holds
the complete processing recipe. Each excerpt is downmixed to mono and its
last 2 seconds are overlapped with its first 2 seconds using complementary
cosine weights, producing a 16-second cyclic loop. A cyclic fourth-order
spectral high-pass removes rumble below 40 Hz for the river, 45 Hz for the
lake, and 28 Hz for the waterfall. The loop's phase is rotated to place a
quiet adjacent sample pair at the codec boundary. The signal is normalised to a -3 dBFS
peak, encoded as Ogg Vorbis at FFmpeg quality 4, then decoded and checked;
the encoder input is reduced if necessary to keep the decoded peak at or
below -3 dBFS. Native preview sample rates are retained: 44.1 kHz for the
river and 48 kHz for the lake and waterfall.

The river recording contains flowing/babbling water, the lake recording
contains waves lapping on a lake shore, and the waterfall recording is a
layered Niagara Falls recording with a deeper continuous roar. No music
or generated sound is mixed into these assets.