Skip to content

Audio Transcribers

AudioTranscriber

Bases: Protocol

A protocol defining the interface for audio transcription services.

This protocol specifies the contract that all audio transcribers in the Mosaico project should adhere to. It defines a single method, :meth:transcribe, which takes an audio asset and returns a transcription.

Implementations of this protocol can use various transcription technologies, such as speech-to-text APIs, local models, or custom algorithms. The protocol ensures a consistent interface regardless of the underlying implementation.

Note

This is a runtime checkable protocol, which means isinstance() and issubclass() checks can be performed against it.

Example:

    class MyTranscriber:
        def transcribe(self, audio_asset: AudioAsset) -> Transcription:
            # Implement transcription logic here
            ...

    transcriber: AudioTranscriber = MyTranscriber()
    transcription = transcriber.transcribe(my_audio_asset)

transcribe 

transcribe(audio_asset: AudioAsset) -> Transcription

Transcribe speech from an audio asset to text.

This method should implement the core logic for converting speech in the provided audio asset into text. The specific implementation can vary based on the transcription technology being used.

Parameters:

Name Type Description Default
audio_asset AudioAsset

The audio asset containing the speech to be transcribed.

 required

Returns:

Type Description
Transcription

A Transcription object containing the text transcription of the speech. .. note:: The implementation should handle various audio formats and durations as defined by the :class:AudioAsset class. It should also be able to handle potential errors gracefully.
Source code in src/mosaico/audio_transcribers/protocol.py 
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
def transcribe(self, audio_asset: AudioAsset) -> Transcription:
    """
    Transcribe speech from an audio asset to text.

    This method should implement the core logic for converting speech in the
    provided audio asset into text. The specific implementation can vary
    based on the transcription technology being used.

    :param audio_asset: The audio asset containing the speech to be transcribed.
    :return: A Transcription object containing the text transcription of the speech.

    .. note::
       The implementation should handle various audio formats and durations
       as defined by the :class:`AudioAsset` class. It should also be able to handle
       potential errors gracefully.
    """
    ...