assets

AKTSER Test Assets

This directory contains audio files used for developing and testing AKTSER, the Telegram bot that summarizes voice messages.

Files

Original Test Audio

• test0.wav
• test1.wav
• test2.ogg

These files were sourced from YouTube videos to simulate real-world voice messages with varying lengths, speech patterns, and background noise.

Processed Audio

• test0_cut.wav
• test1_cut.wav
• test2_cut.wav

These are the outputs of AKTSER's audio trimming feature, which removes silence and low-volume segments from the original files. They demonstrate how AKTSER can reduce message length without losing content.

Audio Sources

thank to creators of the following videos for providing diverse audio content that helped shape AKTSER's capabilities:

1. test0.wav:

   • Duration: 18s
   • Description: a AI generated voice message.
   • Trimming Result: 11s (39% reduction)

2. test1.wav: Youtube

   • Duration: 48s
   • Description: random english talk.
   • Trimming Result: 44s (8% reduction)

3. test2.wav: Youtube

   • Duration: 57s
   • Description: arabic talk.
   • Trimming Result: 51s (11% reduction)

Usage in Development

1. Algorithm Testing: Used to fine-tune our silence detection and trimming algorithms.
2. Accuracy Verification: Manually verified that trimmed files retain all important content.
3. Edge Cases: Helped identify challenges like distinguishing between intentional pauses and unnecessary silence.

Legal Note

These audio files are used under Fair Use for educational and research purposes. If you're a content owner and have concerns, please contact me

Related Jupyter Notebook

For detailed analysis of these audio files, including waveform visualization and statistical analysis, see ../analyze.ipynb.

---

AKTSER: Summarizing voices, saving time. 🎙️⏱️