This promptfoo example evaluates ElevenLabs Forced Alignment, which takes an audio file and its transcript and produces time-aligned output - either a JSON structure of aligned words or SRT-format subtitles. Two providers, elevenlabs:alignment:json and elevenlabs:alignment:srt, expose the two output formats, and each test case supplies an audioFile and transcript pair to be aligned.

Use this example when you need to generate subtitles or word-level timing data from an existing audio recording and its known transcript - for instance, subtitle generation for video content where the spoken text is already known. It is an alignment tool, not a transcription tool: it requires you to already have the transcript text; if you only have audio and need the transcript generated first, this is not the right building block on its own.

The shared prompt is '{{transcript}}', with audioFile and transcript passed as test variables (a format var also appears in the test data). A defaultTest block asserts not-contains: error on every test. Format-specific tests add their own checks - the JSON provider's output is checked for a words key, and the SRT provider's output is checked for SRT timing markers and matching text:

providers:
  # Basic alignment (JSON output)
  - id: elevenlabs:alignment:json
    label: Alignment (JSON)

  # SRT subtitle format
  - id: elevenlabs:alignment:srt
    label: Alignment (SRT Subtitles)

tests:
  - description: Align Armstrong to SRT format
    vars:
      audioFile: examples/elevenlabs-stt/audio/sample1.mp3
      transcript: "That's one small step for man, one giant leap for mankind."
      format: srt
    assert:
      - type: javascript
        value: output.includes('-->') && output.includes('small step')

Runs against ElevenLabs' alignment providers (elevenlabs:alignment:json and elevenlabs:alignment:srt) through promptfoo, and uses sample audio files from the companion elevenlabs-stt example.

Developers building subtitle-generation or transcript-to-audio-sync pipelines who want to validate ElevenLabs Forced Alignment output in either JSON or SRT format before wiring it into a production video or captioning workflow.