Transkript

Ein Transkript ist eine Reduzierung gesprochenen Wortes auf eine schriftliche Form. Dabei geht Kontext zum Teil verloren.

Auto-Transcript

Whisper

→ Hauptartikel: Whisper (OpenAI)

Steps:

1. install Python 3.9.13
   1. setup pip (systemvariables) to go to 3.9.13
2. install pytorch via pip
3. install whsiper via pip
   1. pip install git+https://github.com/openai/whisper.git

https://github.com/openai/whisper/discussions/categories/show-and-tell

• To Wiki: https://www.mediawiki.org/wiki/Extension:OAuth/de

Word

300 minuten / Monat max.

Word to Wiki convert

Regex

Find:

0
Replace:

|-
| 0
Find:
Speaker
Replace:
||
Find:
Transcript
Replace:
{| class="wikitable"
|+ Transcript
|-
! Timestamp !! Speaker !! Text
Find:



Replace:
||

Find:
||
|-
Replace:

|-
Find:
||

Replace:

|}
Find(enable regex):
(?<=\d\d) \|\|\n
Replace:
|| ||

Find(enable regex):
^Audio file \|\|$
.+
\|\}\|\|
\{\|
Replace:
{|
#Done.

Quellen

stackoverflow: better way to automate mouse&keyboard using pyautogui

Tool: pywinauto

Sleep vs tight-loop: https://stackoverflow.com/questions/34361728/windows-desktop-gui-automation-using-python-sleep-vs-tight-loop

Word-Alternativen

AssemblyAI https://www.youtube.com/watch?v=rTWM5WuPhlQ

pydub, ffmpeg and SpeechRecognition https://pythonbasics.org/transcribe-audio/

Automatic Speech Recognition (ASR) - German https://github.com/AASHISHAG/asr-german

Whisper https://openai.com/blog/whisper/