Experimental: Contact Us
This surface is directional and may change before public release. Contact Spiral if you want to use it in an early workflow.
The Audio Extension is intended for speech, music, and event-audio datasets where users need clip metadata, temporal windows, transcripts, features, and deferred waveform materialization.
Planned table functions
| Function | Purpose |
|---|---|
audio.scan(path) | Discover audio files or read a manifest. |
audio.windows(path, window, stride) | Produce fixed-length deferred audio windows. |
audio.transcripts(path) | Read transcript rows aligned to clips or windows. |
audio.features(audio_ref) | Produce preview feature rows such as duration or sample rate. |
audio.decode(audio_ref) | Materialize waveform windows. |
These names are preview syntax. The functions are not registered in the current default CLI session.
Output shape
Window functions should produce rows like:
| Column | Meaning |
|---|---|
clip_id | Stable clip id. |
path | Local path or object URI. |
start_ms | Window start time. |
duration_ms | Window duration. |
sample_rate | Sample rate when known. |
label | Optional label. |
audio_ref | Deferred audio window reference. |
Example shape
SELECT clip_id, window_ref, label
FROM audio.windows('s3://bucket/audio/', window => '10s', stride => '5s');Notes
- Metadata queries should avoid waveform decode.
- Transcript joins should use stable clip/window ids.
- Materialization options should document sample rate, channel layout, waveform dtype, spectrogram policy, and device placement once implemented.
Last updated on