{
  "id": "qwen3_asr_subtitles",
  "description": "Generate transcript, word timestamps, and SRT subtitles from speech using Qwen3 ASR with model-side audio chunking.",
  "inputs": {
    "source_audio": "${workflow_dir}/../asr_validation/librispeech/librispeech_test_clean_6930-75918-0000.wav",
    "language": "en",
    "qwen3_asr_model": "",
    "qwen3_forced_aligner_model": ""
  },
  "steps": [
    {
      "id": "prepare_source",
      "type": "convert_audio",
      "input": "${source_audio}",
      "output": "${out_dir}/source/source_16k.wav",
      "sample_rate": 16000,
      "channels": 1
    },
    {
      "id": "transcribe",
      "type": "model",
      "family": "qwen3_asr",
      "model": "${qwen3_asr_model}",
      "task": "asr",
      "mode": "offline",
      "text_out": "${out_dir}/transcript.txt",
      "words_out": "${out_dir}/words.json",
      "session_options": {
        "qwen3_asr.forced_aligner_model_path": "${qwen3_forced_aligner_model}"
      },
      "request": {
        "id": "transcribe",
        "audio": "${prepare_source.audio_path}",
        "language": "${language}",
        "text": "Transcribe the speech exactly.",
        "audio_chunk_mode": "auto"
      }
    },
    {
      "id": "format_srt",
      "type": "format_subtitles",
      "format": "srt",
      "words": "${transcribe.words_path}",
      "output": "${out_dir}/subtitles.srt",
      "sample_rate": 16000
    }
  ]
}