trax/tests/fixtures/README.md

Test Fixtures

This directory contains test fixtures for the Trax comprehensive testing suite.

Audio Files Required

Place the following real audio files in the audio/ directory for comprehensive testing:

Required Test Audio Files

1. sample_5s.wav - 5-second clear speech audio (16kHz mono WAV)

   • Used for quick transcription tests
   • Should contain clear, simple speech

2. sample_30s.mp3 - 30-second audio with varied content (MP3 format)

   • Used for format conversion testing
   • Should include some technical terms

3. sample_2m.mp4 - 2-minute video file with audio (MP4 format)

   • Used for video processing pipeline tests
   • Should contain continuous speech

4. sample_noisy.wav - Audio with background noise

   • Used for noise handling tests
   • Should contain speech with background noise

5. sample_multi.wav - Multi-speaker audio

   • Used for speaker diarization tests (v4)
   • Should contain multiple distinct speakers

6. sample_tech.mp3 - Technical content audio

   • Used for technical vocabulary testing
   • Should contain technical terms, acronyms, etc.

Usage

These files are automatically detected by the test fixtures in conftest.py and used throughout the test suite.

Important: The comprehensive testing suite follows the project rule of using real audio files instead of mocks to catch real-world edge cases.

Creating Test Files

If you need to create test audio files for development:

# Generate a 5-second test tone (requires ffmpeg)
ffmpeg -f lavfi -i "sine=frequency=440:duration=5" -ar 16000 -ac 1 sample_5s.wav

# Convert to different formats
ffmpeg -i sample_5s.wav -ar 44100 sample_30s.mp3
ffmpeg -i sample_5s.wav -c:v libx264 -c:a aac sample_2m.mp4

Performance Requirements

Test files should enable verification of these v1 pipeline requirements:

• 5-minute audio processed in <30 seconds
• Memory usage <2GB
• 95% transcription accuracy on clear audio
• Real-time factor <0.1 (much faster than real-time)