Amondin

A simple and private transcription tool able to segment speakers and convert audio to text.

Features

• convert audio to text
• segment speakers
• supporting multiple file formats (.wav is guaranteed to work)
• detecting number of speakers (hard setting improves performance)
• detecting language (hard setting improves performance)
• return transcribed dialogs as .csv

Installation

pip install git+https://github.com/tim-roethig-db/amondin.git@master

Usage

from amondin import transcribe

transcribe(
    "sample.m4a", "sample.csv",
    hf_token="<your hf token>",
    s2t_model="openai/whisper-large-v3",
    device="cuda",
    language="german",
    num_speakers=2
)