izwi align

Forced alignment — align text to audio at word level.

Synopsis

izwi align <FILE> <TEXT> [OPTIONS]

Description

Aligns reference text to audio, producing word-level timestamps. Useful for:

• Subtitle generation
• Karaoke timing
• Audio editing
• Pronunciation analysis

Arguments

Options

Examples

Basic alignment

izwi align audio.wav "Hello world, this is a test."

Save to file

izwi align audio.wav "Hello world" --output alignment.json

Text output

izwi align audio.wav "Hello world" --format text

Output Formats

JSON (default)

{
  "alignments": [
    {"word": "Hello", "start": 0.0, "end": 0.45},
    {"word": "world", "start": 0.50, "end": 0.95},
    {"word": "this", "start": 1.10, "end": 1.30},
    {"word": "is", "start": 1.35, "end": 1.45},
    {"word": "a", "start": 1.50, "end": 1.55},
    {"word": "test", "start": 1.60, "end": 2.00}
  ],
  "duration": 2.0
}

Text

Hello     0.00 - 0.45 world     0.50 - 0.95 this      1.10 - 1.30 is        1.35 - 1.45 a         1.50 - 1.55 test      1.60 - 2.00

Use Cases

Subtitle Generation

Generate precise timestamps for subtitles:

izwi align video_audio.wav "$(cat script.txt)" --output subtitles.json

Audio Editing

Find exact word boundaries for editing:

izwi align podcast.wav "um actually" --format json

Pronunciation Analysis

Analyze timing of spoken words:

izwi align recording.wav "The quick brown fox" --format verbose_json

Available Models

See Also

• izwi transcribe — Speech-to-text
• izwi diarize — Speaker diarization