conversallink 0.4.2

Lightweight CLI for Recording & Transcribing with Whisper Accuracy

ConversalLink

Conversation Recorder & Transcriber

---

🎯 Features

• 🎙️ Record audio in 16kHz mono WAV format
• ✍️ Take timestamped notes while recording
• 📂 Use existing audio files (WAV, MP3, FLAC, OGG…)
• 🧠 Transcribe using Faster-Whisper (select model: tiny → large)
• 🔗 Auto-align notes with transcript segments
• 📤 Export as:
  • Markdown
  • JSON
  • Plain text
  • Beautiful HTML (with dark mode, audio player, jump-to-note)

---

📦 Installation

From PyPI (Recommended)

pip install conversallink

From Source

• Clone the repository

git clone https://github.com/gchatzigianniss/ConversalLink

• Install dependencies

pip install -r requirements.txt

• Run the script

python conversallink.py --file conversation.wav --notes conversation_notes.json

---

🚀 Quick Usage Examples

1. Record + Take Notes

# Run as a command-line tool
conversallink --record conversation.wav --model small --out markdown

# Or using Python module syntax
python -m conversallink --record conversation.wav --model small --out markdown

2. Transcribe an Existing File + Merge Notes

python conversallink.py --file conversation.wav --notes conversation_notes.json

3. Transcribe Only

python conversallink.py --file podcast.mp3 --model medium --out html

🤖 Transcription Models

ConversalLink uses Faster-Whisper models for transcription. You can select the model size based on your needs for accuracy vs. speed and resource usage.

Model Size	Parameters	Accuracy	Speed	VRAM Usage	Disk Space	Use Case
tiny	39M	Low	Very Fast	~1GB	~80MB	Quick drafts, low-resource environments
base	74M	Basic	Fast	~1GB	~150MB	Simple recordings, clear speech
small	244M	Good	Medium	~1GB	~500MB	General purpose, good balance
medium	769M	Very Good	Medium-Slow	~2GB	~1.5GB	Professional transcripts
large-v2	1550M	Excellent	Slow	~3GB	~3GB	Highest accuracy needs
large-v3	1550M	Best	Slow	~3GB	~3GB	Latest model with best results

To select a model, use the --model parameter:

conversallink --file recording.wav --model medium

Note: Larger models provide better accuracy but require more computational resources and run slower. The small model offers a good balance for most use cases.

---

---

📦 Dependencies

Install everything with:

pip install -r requirements.txt

Core:

• sounddevice
• soundfile
• faster-whisper
• rich

Optional (for non-WAV input):

• pydub
• ffmpeg (must be available in your system PATH)

---

✅ Output Files

conversation_transcript.md     # 📝 Markdown
conversation_transcript.json   # 🗄️ JSON (full data)
conversation_transcript.html   # 🌐 Interactive HTML player
conversation_transcript.txt    # 📄 Plain text
conversation_notes.json        # 🧷 Timestamped notes

---

📄 License

This project is open source under the MIT License.

---

👨‍💻 Author

gchatzigianniss