https://github.com/bskinn/aai-transcribe

AssemblyAI diarized transcriber script
https://github.com/bskinn/aai-transcribe

Last synced: 21 days ago
JSON representation

AssemblyAI diarized transcriber script

• Host: GitHub
• URL: https://github.com/bskinn/aai-transcribe
• Owner: bskinn
• License: mit
• Created: 2023-06-26T21:53:12.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2024-02-06T21:49:20.000Z (10 months ago)
• Last Synced: 2024-05-01T17:33:36.090Z (7 months ago)
• Language: Python
• Homepage:
• Size: 6.84 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE.txt

Awesome Lists containing this project

README

        
# aai-transcribe

Helper repo for submitting an audio file to AssemblyAI for

diarized transcription.

**NOTE:** I don't have any plans for further development on this,

or to keep it updated as Python and the AssemblyAI API evolves,

so I'm archiving it.

## Setup

Clone the repo:

```

$ git clone https://github.com/bskinn/aai-transcribe

```

Enter the cloned repo directory, create and activate a Python 3.11+ virtualenv.

Update `pip`:

```

(env) $ python -m pip install -U pip

```

Install dependencies:

```

(env) $ python -m pip install -r requirements.txt

```

Obtain an AssemblyAI API token and store it as the `AAI_TOKEN`

environment variable.

## Transcription

Activate the virtual environment and pass the path to the audio file to be

transcribed into the `aai.py` script:

```

(env) $ python aai.py {path to audio file}

```

The script will handle the initial submission to AssemblyAI, followed by

the request for diarized transcription. It then polls the AssemblyAI

status endpoint until transcription has completed, then downloads the

resulting transcript JSON payload as `utterances.json`.

It then produces two files:

- `transcript_raw.txt` -- The transcript without diarization

- `transcript_split.txt` -- The transcript with diarization

AssemblyAI does not have access to the names of the speakers, so the script outputs

anonymized labels for the transcription instead (e.g., `[[A]]` and `[[B]]`). It's thus

necessary to manually do a find-and-replace to swap in each speaker's name in place

of their name placeholder.

The speaker identification is usually fairly reliable with two or three unique speakers.

With larger groups, inaccuracies are common. Manual cross-checking of all speaker names

is a good idea.