Audio To Text

Defines a base class for audio to text converter used in Gen AI applications.

Quick Start

# Assume `converter` is an instance of a BaseAudioToText subclass

# Returns a TextResult object with combined text and metadata
result = await converter.convert_to_text_result(source="path/to/audio.mp3")

# Or for backward compatibility, returns a list of AudioTranscript segments
segments = await converter.convert(audio_source="path/to/audio.mp3")

`BaseAudioToText()`

Bases: BaseModalityConverter, Component, ABC

An abstract base class for audio to text used in Gen AI applications.

Extends BaseModalityConverter so audio converters participate in the shared modality-converter hierarchy. The primary convert API remains list-based for backward compatibility; use convert_to_text_result for the unified TextResult contract.

`convert(audio_source=None, *, source=None, **kwargs)` `async`

Convert audio to timestamped transcript segments.

This is the primary backward-compatible API used across gllm-* integrations.

Parameters:

Name	Type	Description	Default
`audio_source`	`str \| bytes \| None`	The source of the audio to convert. Accepted as a positional or keyword argument for backward compatibility.	`None`
`source`	`str \| bytes \| None`	Keyword-only alias for `audio_source`, used by the `BaseModalityConverter` call contract.	`None`
`**kwargs`	`Any`	Additional transcription parameters.	`{}`

Returns:

Type	Description
`list[AudioTranscript]`	list[AudioTranscript]: Timestamped transcript segments.

Raises:

Type	Description
`ValueError`	If neither `audio_source` nor `source` is provided.

`convert_to_text_result(audio_source=None, *, source=None, **kwargs)` `async`

Convert audio to the unified TextResult format used by modality converters.