# TIDEPOOL.ING / production and playback

An original fictional sound artwork by Kris Krüg with Codex. Created in 2026, set on an imagined Salish Sea in 2075. Opening edition: Eleven Days Underwater.

## The hour

Seven programmes, exactly 60 minutes. Ninety-two spoken cues total 6 minutes 4.013 seconds. Between 26:00 and 36:00, Open Water contains ten uninterrupted minutes of composed synthetic coastal sound, without speech or pitched music.

All dialogue, characters, sponsors, businesses and future conditions are fiction. Stock Kokoro v1.0 voices: af_heart, bm_george, bf_emma, af_sarah, am_puck, af_bella, am_michael, am_fenrir. The model is Apache-2.0 licensed and the local kokoro-onnx runtime is MIT licensed. No voice cloning is used. Music, water and interference are original procedural synthesis. No real shoreline recording, archival voice or listener message is included. The relay illustrations are generated concept art.

The downloadable master is 48 kHz stereo, 24-bit PCM WAV. Web listening uses 320 kbps MP3. The public remaster has been checked for clipping, programme duration and mono compatibility. Complete headphone and speaker listening review remains an artistic checkpoint. Music and water have intentional dynamic range; start at a comfortable volume.

## How the shared broadcast works

The station plays an authored recording on a repeating schedule. It is not a live performance or a physical radio-frequency broadcast. Each receiver calculates its place in a sixty-minute cycle from the UTC anchor 3 October 2026 at 00:00:00. Normally synchronized device clocks let listeners join the same moment. An incorrectly set device clock can put a listener ahead or behind.

Nothing plays until you choose Join Transmission. Stop silences the receiver. Joining again returns to the scheduled position. Archive playback has its own seekable position; Return to Broadcast rejoins the shared hour. The receiver prepares the next cycle near the hour boundary. Network delays and browser background suspension can interrupt playback.

Reception tuning adds a local listening effect. The visual trace responds to playing audio. Neither shows measured environmental conditions. Reduce Signal Motion is available in Field Notes, and the receiver also respects the system reduced-motion preference.

## Voice provenance

Model: https://huggingface.co/hexgrad/Kokoro-82M
Runtime: https://github.com/thewh1teagle/kokoro-onnx

The public recordings use a replacement stock cast. Original system-voice listening mixes are preserved privately. Scripts, programme order and original synthesis are retained. No model weights or software binaries are distributed in the station packs.

## Archive and source notes

The complete station pack contains 31 MP3s: the opening hour, seven standalone programmes, the original fourteen-minute pilot and its twenty-two pieces. It also includes scripts, captions and cue sheets. Individual recordings have lossless WAV downloads. The pilot receiver additionally preserves the earlier 1:58 audition.

This station does not request a microphone, use accounts, collect listener submissions or include an analytics script. Ordinary hosting access logs may be processed by the hosting provider. The fiction's messages and call-ins were authored for the work.