tatt

Transcribe All The Things™

tatt is a CLI for creating and managing speech-to-text transcripts.

Installation

pip install tatt

Dependencies

1. A recording to transcribe.
2. a) An AWS account or b) a Google Cloud account with the speech-to-text API and Cloud Storage enabled.
3. ffmpeg for some operations, but tatt will let you know when it's needed.

Usage

List All Commands

$ transcribe --help

Usage: transcribe [OPTIONS] COMMAND [ARGS]...

Options:
  --help  Show this message and exit.

Commands:
  get        Downloads and/or saves completed transcript.
  languages
  list       Lists available STT services.
  services   Lists available speech-to-text services.
  status     Check the status of a transcription job.
  this       Sends a media file to be transcribed.

List All STT Services

$ transcribe services

Here are all the available speech-to-text services:

  amazon		$0.006 per 15 seconds
  google		$0.009 per 15 seconds

Get Something Transcribed

$ transcribe this <path_to_media_file> <service_name>

Okay, transcribing <path_to_media_file> using <service_name>...
Okay, job <job_name> is being transcribed.  Use "get" command to download it.

$ transcribe this --help

Usage: transcribe this [OPTIONS] MEDIA_FILEPATH SERVICE_NAME

Sends a media file to be transcribed.

Options:
  --punctuation                   only for Google Speech, defaults to True
  --speaker-id / --no-speaker-id  only for google and amazon, defaults to True
  --num_speakers INTEGER          only for google and amazon, defaults to 2
  --model TEXT                    only for Google Speech, defaults to
                                  "phone_call"
  --use-enhanced                  only for Google Speech, defaults to True
  --language-code TEXT            only for google and amazon, defaults to en-
                                  US
  --help                          Show this message and exit.

List Transcripts

$ transcribe list

Service Job Name                                Status
------- --------                                ------
amazon  tester.mp3                              IN_PROGRESS
amazon  messed_up.mp4                           FAILED
amazon  done_test.mp3                           COMPLETED
amazon  also_done.MP3                           COMPLETED
google  hey_there.mp3                           COMPLETED


$ transcribe list <job_name>

Service Job Name                                Status
------- --------                                ------
amazon  <job_name>                              IN_PROGRESS

Get A Completed Transcript

$ transcribe get <job_name> # prints to stdout

{'accountId': '416321668733',
 'jobName': 'a1bace5e-8b08-4ce4-b08c-834a23aafcf1',
 'results': {'items': [{'alternatives': [{'confidence': '0.9774',
                                          'content': 'Hi'}],
                        'end_time': '1.5',
                        'start_time': '1.23',
                        'type': 'pronunciation'},
                       {'alternatives': [{'confidence': '0.9429',
                                          'content': 'is'}],
                        'end_time': '1.71',
                        'start_time': '1.5',
                        'type': 'pronunciation'},
                       ...

$ transcribe get --save <job_name>

Saved transcript to <job_name>.json

Services Supported

• Amazon Transcribe
• Google Cloud Speech-to-Text

Planned

• Watson
• Kaldi and/or things built on it
• Speechmatics
• Mozilla's new open-source STT thing