TranscripService

Local video transcription using OpenAI Whisper — no API key, no cloud, no cost.

---

Features

• Transcribe MP4 and other video formats to plain text, fully offline
• Real timestamps sourced from Whisper's native speech segments ([MM:SS - MM:SS])
• Live progress indicator during transcription
• Batch processing: automatically transcribe all MP4 files in a folder
• Dual output per video: raw transcript + formatted transcript with timestamps
• Skip logic: batch script skips files already transcribed

---

How It Works

Video file (.mp4)
      │
      ▼
 ffprobe (FFmpeg)
  └─ extracts audio duration for progress estimation
      │
      ▼
 OpenAI Whisper (local, 'small' model)
  └─ speech-to-text with native segment timestamps
      │
      ▼
 Two output files:
  ├─ <name>_transcription.txt   (raw continuous text)
  └─ <name>_formatted.txt       (timestamped segments)

Whisper runs entirely on your machine — no audio is sent to any external service. The model (~461 MB) is downloaded automatically on first run.

---

Prerequisites

• Python 3.8+
• FFmpeg installed and available in your system PATH
  • Windows: ffmpeg.org/download or winget install ffmpeg
  • macOS: brew install ffmpeg
  • Linux: sudo apt install ffmpeg
• ~500 MB free disk space (Whisper small model, downloaded once on first run)

---

Installation

git clone https://github.com/juancho-cl/TranscripService.git
cd TranscripService
pip install -r requirements.txt

---

Usage

Single file — basic

python transcribe.py "meeting_recording.mp4"

Output: meeting_recording_transcription.txt

---

Single file — with progress bar

python transcribe_with_progress.py "meeting_recording.mp4"

Output:

• meeting_recording_transcription.txt
• meeting_recording_formatted.txt

---

Single file — with real timestamps (recommended)

python transcribe_with_segments.py "meeting_recording.mp4"

Output:

• meeting_recording_transcription.txt
• meeting_recording_formatted.txt

---

Batch processing (Windows)

Place Script.bat and all your .mp4 files in the same folder, then run:

Script.bat

The script will:

1. Process every .mp4 in the folder
2. Skip files already transcribed
3. Save outputs to output/
4. Save logs to logs/

---

Script Comparison

Script	Output files	Timestamps	Best for
transcribe.py	1 (raw text)	None	Quick transcription, minimal setup
transcribe_with_progress.py	2 (raw + formatted)	Simulated	Visual feedback during long files
transcribe_with_segments.py	2 (raw + formatted)	Real (Whisper native)	Accurate timestamped transcripts

---

Sample Output

[00:02 - 00:07]
So this is the current server and this is a shared folder.

[00:09 - 00:15]
Right now you can see there are two different folders here.

[00:15 - 00:22]
What I'm going to show you is how to map this as a network drive on your computer.

[00:22 - 00:28]
Once it's mapped, it will appear just like any other drive in File Explorer.

---

License

MIT — see LICENSE.