Audio#

Audio Operations#

Information#

Audio information, validation, and filter operators.

These operators inspect audio objects and return scalar values or row values. They do not modify waveform content. Encoded inputs are accepted only at the pipeline boundary; waveform-only operators require an explicit audio_decode step so that pipeline definitions stay predictable.

is_valid_audio(*columns[, validation, ...])

Check whether an audio input looks like valid audio.

audio_metadata(*columns[, concurrency])

Extract audio metadata.

audio_duration(*columns[, concurrency])

Return audio duration in seconds.

audio_duration_filter(*columns[, ...])

Check whether audio duration is within optional second bounds.

audio_size_filter(*columns[, min_bytes, ...])

Check whether encoded audio size is within optional byte bounds.

audio_silence_detection(*columns, threshold_db)

Detect low-amplitude silence ranges in an AUDIO_WAVEFORM.

audio_detect_speech(*columns[, ...])

Detect speech activity ranges in a mono waveform with WebRTC VAD.

Transform#

Audio decode, encode, split, and waveform transform operators.

The transform layer separates encoded boundary inputs from decoded waveform processing. Operators that modify samples accept AUDIO_WAVEFORM only, so a pipeline should usually call audio_decode once, transform waveforms, and then call audio_encode or a sink-specific writer at the boundary.

audio_decode(*columns[, on_error, ...])

Decode audio into a waveform row.

audio_encode(*columns[, format, concurrency])

Encode AUDIO_WAVEFORM to bytes.

audio_convert_format(*columns[, format, ...])

Convert encoded audio to another encoded format.

audio_standardize(*columns[, sample_rate, ...])

Standardize an audio waveform.

audio_resample(*columns, sample_rate[, ...])

Resample an audio waveform.

audio_split_by_duration(*columns, ...[, ...])

Split audio by fixed duration into one output row per segment.

audio_split_by_timestamp(*columns[, ...])

Split audio by explicit timestamp ranges into one output row per segment.

audio_split_by_speech(*columns[, ...])

Split audio by speech activity ranges into one output row per segment.

audio_concat(*columns[, concurrency])

Concatenate waveforms.

Speech#

Speech model operators for decoded audio waveforms.

These operators are model-backed pandas UDFs. They accept standardized AUDIO_WAVEFORM values only: callers should decode boundary inputs and normalize to 16 kHz mono before invoking Whisper-based operators.

audio_asr_whisper(*columns[, language, ...])

Run Whisper automatic speech recognition over audio waveforms.

audio_detect_language(*columns[, top_k, ...])

Detect likely spoken languages in an audio waveform.