{ "success": true, "input_file": "demo_audio\\Yuri_Kizaki.mp3", "audio_metadata": { "duration_seconds": 32.366, "sample_rate": 44100, "channels": 1, "sample_width": 2, "frame_count": 1427328.0, "max_possible_amplitude": 32768.0 }, "processing_stats": { "total_time": 131.9166796207428, "component_times": { "audio_preprocessing": 7.074368000030518, "speaker_diarization": 19.895120859146118, "speech_recognition": 51.43702697753906, "translation": 6.94795036315918, "output_formatting": 0.0 }, "num_speakers": 1, "num_segments": 12, "languages_detected": [ "ja" ], "total_speech_duration": 26.021250000000002 }, "outputs": { "json": "{\n \"metadata\": {\n \"audio_filename\": \"Yuri_Kizaki.mp3\",\n \"processing_timestamp\": \"2025-09-02T16:18:58.085380\",\n \"total_segments\": 12,\n \"total_speakers\": 1,\n \"languages_detected\": [\n \"ja\"\n ],\n \"total_audio_duration\": 31.90784375,\n \"total_speech_duration\": 26.021250000000002,\n \"speech_ratio\": 0.8155126433449456,\n \"audio_metadata\": {\n \"duration_seconds\": 32.366,\n \"sample_rate\": 44100,\n \"channels\": 1,\n \"sample_width\": 2,\n \"frame_count\": 1427328.0,\n \"max_possible_amplitude\": 32768.0\n },\n \"processing_stats\": {\n \"audio_preprocessing\": 7.074368000030518,\n \"speaker_diarization\": 19.895120859146118,\n \"speech_recognition\": 51.43702697753906,\n \"translation\": 6.94795036315918\n }\n },\n \"statistics\": {\n \"total_duration\": 31.90784375,\n \"total_speech_duration\": 26.021250000000002,\n \"speech_ratio\": 0.8155126433449456,\n \"average_segment_duration\": 2.1684375,\n \"longest_segment\": 5.248125000000002,\n \"shortest_segment\": 0.354375000000001,\n \"average_confidence_diarization\": 1.0,\n \"average_confidence_transcription\": -0.27468773681238773,\n \"average_confidence_translation\": 0.7999999999999999,\n \"total_words_original\": 12,\n \"total_words_translated\": 75\n },\n \"segments\": [\n {\n \"start_time\": 0.40221875,\n \"end_time\": 4.77284375,\n \"speaker_id\": \"SPEAKER_00\",\n \"original_text\": \"音声メッセージが既存のウェブサイトを超えたコミュニケーションを実現。\",\n \"original_language\": \"ja\",\n \"translated_text\": \"The audio message will bring out communication beyond the existing website.\",\n \"confidence_diarization\": 1.0,\n \"confidence_transcription\": -0.1825541319946448,\n \"confidence_translation\": 0.8,\n \"word_timestamps\": [\n {\n \"word\": \"音\",\n \"start\": 0.40221875,\n \"end\": 0.56221875,\n \"confidence\": 0.8530172109603882\n },\n {\n \"word\": \"声\",\n \"start\": 0.56221875,\n \"end\": 0.80221875,\n \"confidence\": 0.9917272329330444\n },\n {\n \"word\": \"メ\",\n \"start\": 0.80221875,\n \"end\": 0.9422187500000001,\n \"confidence\": 0.9574464559555054\n },\n {\n \"word\": \"ッ\",\n \"start\": 0.9422187500000001,\n \"end\": 1.02221875,\n \"confidence\": 0.999119222164154\n },\n {\n \"word\": \"セ\",\n \"start\": 1.02221875,\n \"end\": 1.14221875,\n \"confidence\": 0.99460768699646\n },\n {\n \"word\": \"ージ\",\n \"start\": 1.14221875,\n \"end\": 1.30221875,\n \"confidence\": 0.9997381567955017\n },\n {\n \"word\": \"が\",\n \"start\": 1.30221875,\n \"end\": 1.5222187500000002,\n \"confidence\": 0.9662947654724121\n },\n {\n \"word\": \"既\",\n \"start\": 1.5222187500000002,\n \"end\": 1.92221875,\n \"confidence\": 0.7296531945466995\n },\n {\n \"word\": \"存\",\n \"start\": 1.92221875,\n \"end\": 2.08221875,\n \"confidence\": 0.9589823484420776\n },\n {\n \"word\": \"の\",\n \"start\": 2.08221875,\n \"end\": 2.20221875,\n \"confidence\": 0.9912187457084656\n },\n {\n \"word\": \"ウ\",\n \"start\": 2.20221875,\n \"end\": 2.3022187499999998,\n \"confidence\": 0.6959699988365173\n },\n {\n \"word\": \"ェ\",\n \"start\": 2.3022187499999998,\n \"end\": 2.36221875,\n \"confidence\": 0.9874258041381836\n },\n {\n \"word\": \"ブ\",\n \"start\": 2.36221875,\n \"end\": 2.48221875,\n \"confidence\": 0.9893200397491455\n },\n {\n \"word\": \"サ\",\n \"start\": 2.48221875,\n \"end\": 2.64221875,\n \"confidence\": 0.9838968515396118\n },\n {\n \"word\": \"イ\",\n \"start\": 2.64221875,\n \"end\": 2.7222187499999997,\n \"confidence\": 0.9970263838768005\n },\n {\n \"word\": \"ト\",\n \"start\": 2.7222187499999997,\n \"end\": 2.86221875,\n \"confidence\": 0.9971777200698853\n },\n {\n \"word\": \"を\",\n \"start\": 2.86221875,\n \"end\": 2.94221875,\n \"confidence\": 0.9877551198005676\n },\n {\n \"word\": \"超\",\n \"start\": 2.94221875,\n \"end\": 3.04221875,\n \"confidence\": 0.6848042011260986\n },\n {\n \"word\": \"え\",\n \"start\": 3.04221875,\n \"end\": 3.1822187499999997,\n \"confidence\": 0.9907885193824768\n },\n {\n \"word\": \"た\",\n \"start\": 3.1822187499999997,\n \"end\": 3.2822187499999997,\n \"confidence\": 0.9983263611793518\n },\n {\n \"word\": \"コ\",\n \"start\": 3.2822187499999997,\n \"end\": 3.44221875,\n \"confidence\": 0.9066019058227539\n },\n {\n \"word\": \"ミ\",\n \"start\": 3.44221875,\n \"end\": 3.54221875,\n \"confidence\": 0.9985296726226807\n },\n {\n \"word\": \"ュ\",\n \"start\": 3.54221875,\n \"end\": 3.58221875,\n \"confidence\": 0.9981721639633179\n },\n {\n \"word\": \"ニ\",\n \"start\": 3.58221875,\n \"end\": 3.6622187499999996,\n \"confidence\": 0.9988634586334229\n },\n {\n \"word\": \"ケ\",\n \"start\": 3.6622187499999996,\n \"end\": 3.8222187499999998,\n \"confidence\": 0.9971752166748047\n },\n {\n \"word\": \"ー\",\n \"start\": 3.8222187499999998,\n \"end\": 3.90221875,\n \"confidence\": 0.9970790147781372\n },\n {\n \"word\": \"ショ\",\n \"start\": 3.90221875,\n \"end\": 4.00221875,\n \"confidence\": 0.9993009567260742\n },\n {\n \"word\": \"ン\",\n \"start\": 4.00221875,\n \"end\": 4.1022187500000005,\n \"confidence\": 0.9991468191146851\n },\n {\n \"word\": \"を\",\n \"start\": 4.1022187500000005,\n \"end\": 4.18221875,\n \"confidence\": 0.991553008556366\n },\n {\n \"word\": \"実\",\n \"start\": 4.18221875,\n \"end\": 4.36221875,\n \"confidence\": 0.9924994111061096\n },\n {\n \"word\": \"現。\",\n \"start\": 4.36221875,\n \"end\": 4.6022187500000005,\n \"confidence\": 0.9942215085029602\n }\n ],\n \"model_info\": {\n \"diarization_model\": \"pyannote/speaker-diarization-3.1\",\n \"transcription_model\": \"faster-whisper-small\",\n \"translation_model\": \"google_translate\"\n }\n },\n {\n \"start_time\": 5.5153437499999995,\n \"end_time\": 7.388468750000001,\n \"speaker_id\": \"SPEAKER_00\",\n \"original_text\": \"目で見るだけだったウェブサイトに\",\n \"original_language\": \"ja\",\n \"translated_text\": \"I'm going to show you what I'm doing.\",\n \"confidence_diarization\": 1.0,\n \"confidence_transcription\": -0.22203674035913804,\n \"confidence_translation\": 0.8,\n \"word_timestamps\": [\n {\n \"word\": \"目\",\n \"start\": 5.5153437499999995,\n \"end\": 5.655343749999999,\n \"confidence\": 0.8701557517051697\n },\n {\n \"word\": \"で\",\n \"start\": 5.655343749999999,\n \"end\": 5.815343749999999,\n \"confidence\": 0.991607666015625\n },\n {\n \"word\": \"見\",\n \"start\": 5.815343749999999,\n \"end\": 5.9353437499999995,\n \"confidence\": 0.9280027151107788\n },\n {\n \"word\": \"る\",\n \"start\": 5.9353437499999995,\n \"end\": 6.05534375,\n \"confidence\": 0.9964483976364136\n },\n {\n \"word\": \"だけ\",\n \"start\": 6.05534375,\n \"end\": 6.235343749999999,\n \"confidence\": 0.9943233728408813\n },\n {\n \"word\": \"だ\",\n \"start\": 6.235343749999999,\n \"end\": 6.4353437499999995,\n \"confidence\": 0.9976925849914551\n },\n {\n \"word\": \"った\",\n \"start\": 6.4353437499999995,\n \"end\": 6.57534375,\n \"confidence\": 0.9989917874336243\n },\n {\n \"word\": \"ウ\",\n \"start\": 6.57534375,\n \"end\": 6.67534375,\n \"confidence\": 0.4343600571155548\n },\n {\n \"word\": \"ェ\",\n \"start\": 6.67534375,\n \"end\": 6.735343749999999,\n \"confidence\": 0.9842584133148193\n },\n {\n \"word\": \"ブ\",\n \"start\": 6.735343749999999,\n \"end\": 6.83534375,\n \"confidence\": 0.9933525323867798\n },\n {\n \"word\": \"サ\",\n \"start\": 6.83534375,\n \"end\": 7.0153437499999995,\n \"confidence\": 0.9906386137008667\n },\n {\n \"word\": \"イ\",\n \"start\": 7.0153437499999995,\n \"end\": 7.07534375,\n \"confidence\": 0.9990501999855042\n },\n {\n \"word\": \"ト\",\n \"start\": 7.07534375,\n \"end\": 7.195343749999999,\n \"confidence\": 0.9961349964141846\n },\n {\n \"word\": \"に\",\n \"start\": 7.195343749999999,\n \"end\": 7.315343749999999,\n \"confidence\": 0.989922821521759\n }\n ],\n \"model_info\": {\n \"diarization_model\": \"pyannote/speaker-diarization-3.1\",\n \"transcription_model\": \"faster-whisper-small\",\n \"translation_model\": \"google_translate\"\n }\n },\n {\n \"start_time\": 7.624718750000001,\n \"end_time\": 9.852218750000002,\n \"speaker_id\": \"SPEAKER_00\",\n \"original_text\": \"音声情報をインクルードすることで\",\n \"original_language\": \"ja\",\n \"translated_text\": \"We're going to be able to do that in the next video.\",\n \"confidence_diarization\": 1.0,\n \"confidence_transcription\": -0.2369275689125061,\n \"confidence_translation\": 0.8,\n \"word_timestamps\": [\n {\n \"word\": \"音\",\n \"start\": 7.624718750000001,\n \"end\": 7.7847187500000015,\n \"confidence\": 0.9499445557594299\n },\n {\n \"word\": \"声\",\n \"start\": 7.7847187500000015,\n \"end\": 8.004718750000002,\n \"confidence\": 0.9357801079750061\n },\n {\n \"word\": \"情\",\n \"start\": 8.004718750000002,\n \"end\": 8.164718750000002,\n \"confidence\": 0.9815613627433777\n },\n {\n \"word\": \"報\",\n \"start\": 8.164718750000002,\n \"end\": 8.40471875,\n \"confidence\": 0.9961434602737427\n },\n {\n \"word\": \"を\",\n \"start\": 8.40471875,\n \"end\": 8.544718750000001,\n \"confidence\": 0.992678165435791\n },\n {\n \"word\": \"イ\",\n \"start\": 8.544718750000001,\n \"end\": 8.684718750000002,\n \"confidence\": 0.9322373270988464\n },\n {\n \"word\": \"ン\",\n \"start\": 8.684718750000002,\n \"end\": 8.74471875,\n \"confidence\": 0.9673494696617126\n },\n {\n \"word\": \"ク\",\n \"start\": 8.74471875,\n \"end\": 8.844718750000002,\n \"confidence\": 0.9965403079986572\n },\n {\n \"word\": \"ル\",\n \"start\": 8.844718750000002,\n \"end\": 8.944718750000002,\n \"confidence\": 0.9498746395111084\n },\n {\n \"word\": \"ード\",\n \"start\": 8.944718750000002,\n \"end\": 9.124718750000001,\n \"confidence\": 0.9774163961410522\n },\n {\n \"word\": \"する\",\n \"start\": 9.124718750000001,\n \"end\": 9.364718750000002,\n \"confidence\": 0.9932113885879517\n },\n {\n \"word\": \"こと\",\n \"start\": 9.364718750000002,\n \"end\": 9.56471875,\n \"confidence\": 0.9621437191963196\n },\n {\n \"word\": \"で\",\n \"start\": 9.56471875,\n \"end\": 9.764718750000002,\n \"confidence\": 0.9964655637741089\n }\n ],\n \"model_info\": {\n \"diarization_model\": \"pyannote/speaker-diarization-3.1\",\n \"transcription_model\": \"faster-whisper-small\",\n \"translation_model\": \"google_translate\"\n }\n },\n {\n \"start_time\": 10.274093750000002,\n \"end_time\": 12.31596875,\n \"speaker_id\": \"SPEAKER_00\",\n \"original_text\": \"情報に新しい価値を与え\",\n \"original_language\": \"ja\",\n \"translated_text\": \"And that's what we're going to do.\",\n \"confidence_diarization\": 1.0,\n \"confidence_transcription\": -0.11563345324248075,\n \"confidence_translation\": 0.8,\n \"word_timestamps\": [\n {\n \"word\": \"情\",\n \"start\": 10.274093750000002,\n \"end\": 10.474093750000002,\n \"confidence\": 0.9788916110992432\n },\n {\n \"word\": \"報\",\n \"start\": 10.474093750000002,\n \"end\": 10.694093750000002,\n \"confidence\": 0.9990907907485962\n },\n {\n \"word\": \"に\",\n \"start\": 10.694093750000002,\n \"end\": 10.814093750000001,\n \"confidence\": 0.9892839789390564\n },\n {\n \"word\": \"新\",\n \"start\": 10.814093750000001,\n \"end\": 11.014093750000002,\n \"confidence\": 0.9793343544006348\n },\n {\n \"word\": \"しい\",\n \"start\": 11.014093750000002,\n \"end\": 11.394093750000003,\n \"confidence\": 0.9975306391716003\n },\n {\n \"word\": \"価\",\n \"start\": 11.394093750000003,\n \"end\": 11.574093750000003,\n \"confidence\": 0.981714278459549\n },\n {\n \"word\": \"値\",\n \"start\": 11.574093750000003,\n \"end\": 11.754093750000003,\n \"confidence\": 0.9989857375621796\n },\n {\n \"word\": \"を\",\n \"start\": 11.754093750000003,\n \"end\": 11.854093750000002,\n \"confidence\": 0.9980254173278809\n },\n {\n \"word\": \"与\",\n \"start\": 11.854093750000002,\n \"end\": 12.114093750000002,\n \"confidence\": 0.9476390182971954\n },\n {\n \"word\": \"え\",\n \"start\": 12.114093750000002,\n \"end\": 12.194093750000002,\n \"confidence\": 0.9922704696655273\n }\n ],\n \"model_info\": {\n \"diarization_model\": \"pyannote/speaker-diarization-3.1\",\n \"transcription_model\": \"faster-whisper-small\",\n \"translation_model\": \"google_translate\"\n }\n },\n {\n \"start_time\": 12.36659375,\n \"end_time\": 14.72909375,\n \"speaker_id\": \"SPEAKER_00\",\n \"original_text\": \"他者との差別化に効果を発揮します\",\n \"original_language\": \"ja\",\n \"translated_text\": \"It's not just about being different from other people.\",\n \"confidence_diarization\": 1.0,\n \"confidence_transcription\": -0.2329371053921549,\n \"confidence_translation\": 0.8,\n \"word_timestamps\": [\n {\n \"word\": \"他\",\n \"start\": 12.36659375,\n \"end\": 12.56659375,\n \"confidence\": 0.7133576273918152\n },\n {\n \"word\": \"者\",\n \"start\": 12.56659375,\n \"end\": 12.72659375,\n \"confidence\": 0.594456672668457\n },\n {\n \"word\": \"と\",\n \"start\": 12.72659375,\n \"end\": 12.84659375,\n \"confidence\": 0.9945782423019409\n },\n {\n \"word\": \"の\",\n \"start\": 12.84659375,\n \"end\": 12.96659375,\n \"confidence\": 0.998796820640564\n },\n {\n \"word\": \"差\",\n \"start\": 12.96659375,\n \"end\": 13.10659375,\n \"confidence\": 0.9885448813438416\n },\n {\n \"word\": \"別\",\n \"start\": 13.10659375,\n \"end\": 13.30659375,\n \"confidence\": 0.9973207116127014\n },\n {\n \"word\": \"化\",\n \"start\": 13.30659375,\n \"end\": 13.48659375,\n \"confidence\": 0.9788604378700256\n },\n {\n \"word\": \"に\",\n \"start\": 13.48659375,\n \"end\": 13.60659375,\n \"confidence\": 0.9965766072273254\n },\n {\n \"word\": \"効\",\n \"start\": 13.60659375,\n \"end\": 13.86659375,\n \"confidence\": 0.9582771062850952\n },\n {\n \"word\": \"果\",\n \"start\": 13.86659375,\n \"end\": 14.02659375,\n \"confidence\": 0.9983495473861694\n },\n {\n \"word\": \"を\",\n \"start\": 14.02659375,\n \"end\": 14.12659375,\n \"confidence\": 0.9957448840141296\n },\n {\n \"word\": \"発\",\n \"start\": 14.12659375,\n \"end\": 14.246593749999999,\n \"confidence\": 0.9888325929641724\n },\n {\n \"word\": \"揮\",\n \"start\": 14.246593749999999,\n \"end\": 14.36659375,\n \"confidence\": 0.9894059002399445\n },\n {\n \"word\": \"します\",\n \"start\": 14.36659375,\n \"end\": 14.54659375,\n \"confidence\": 0.9909846782684326\n }\n ],\n \"model_info\": {\n \"diarization_model\": \"pyannote/speaker-diarization-3.1\",\n \"transcription_model\": \"faster-whisper-small\",\n \"translation_model\": \"google_translate\"\n }\n },\n {\n \"start_time\": 15.67409375,\n \"end_time\": 16.06221875,\n \"speaker_id\": \"SPEAKER_00\",\n \"original_text\": \"また!\",\n \"original_language\": \"ja\",\n \"translated_text\": \"Again!\",\n \"confidence_diarization\": 1.0,\n \"confidence_transcription\": -0.4752265453338623,\n \"confidence_translation\": 0.8,\n \"word_timestamps\": [\n {\n \"word\": \"また!\",\n \"start\": 15.67409375,\n \"end\": 15.894093750000001,\n \"confidence\": 0.9813592433929443\n }\n ],\n \"model_info\": {\n \"diarization_model\": \"pyannote/speaker-diarization-3.1\",\n \"transcription_model\": \"faster-whisper-small\",\n \"translation_model\": \"google_translate\"\n }\n },\n {\n \"start_time\": 16.33221875,\n \"end_time\": 21.58034375,\n \"speaker_id\": \"SPEAKER_00\",\n \"original_text\": \"文字やグラフィックだけでは伝えることの難しかった感情やニュアンスを表現し\",\n \"original_language\": \"ja\",\n \"translated_text\": \"It's not just writing, it's graphic.\",\n \"confidence_diarization\": 1.0,\n \"confidence_transcription\": -0.16042621207959723,\n \"confidence_translation\": 0.8,\n \"word_timestamps\": [\n {\n \"word\": \"文\",\n \"start\": 16.33221875,\n \"end\": 16.53221875,\n \"confidence\": 0.8754217624664307\n },\n {\n \"word\": \"字\",\n \"start\": 16.53221875,\n \"end\": 16.69221875,\n \"confidence\": 0.9960361123085022\n },\n {\n \"word\": \"や\",\n \"start\": 16.69221875,\n \"end\": 16.79221875,\n \"confidence\": 0.9906545281410217\n },\n {\n \"word\": \"グ\",\n \"start\": 16.79221875,\n \"end\": 16.892218749999998,\n \"confidence\": 0.9925161004066467\n },\n {\n \"word\": \"ラ\",\n \"start\": 16.892218749999998,\n \"end\": 17.01221875,\n \"confidence\": 0.9981822967529297\n },\n {\n \"word\": \"フ\",\n \"start\": 17.01221875,\n \"end\": 17.072218749999998,\n \"confidence\": 0.9955530762672424\n },\n {\n \"word\": \"ィ\",\n \"start\": 17.072218749999998,\n \"end\": 17.15221875,\n \"confidence\": 0.9970651268959045\n },\n {\n \"word\": \"ック\",\n \"start\": 17.15221875,\n \"end\": 17.27221875,\n \"confidence\": 0.9935983419418335\n },\n {\n \"word\": \"だけ\",\n \"start\": 17.27221875,\n \"end\": 17.45221875,\n \"confidence\": 0.9928644895553589\n },\n {\n \"word\": \"では\",\n \"start\": 17.45221875,\n \"end\": 17.67221875,\n \"confidence\": 0.9097373485565186\n },\n {\n \"word\": \"伝\",\n \"start\": 17.67221875,\n \"end\": 17.91221875,\n \"confidence\": 0.9866331815719604\n },\n {\n \"word\": \"える\",\n \"start\": 17.91221875,\n \"end\": 18.09221875,\n \"confidence\": 0.9961875081062317\n },\n {\n \"word\": \"こと\",\n \"start\": 18.09221875,\n \"end\": 18.232218749999998,\n \"confidence\": 0.8297985792160034\n },\n {\n \"word\": \"の\",\n \"start\": 18.232218749999998,\n \"end\": 18.43221875,\n \"confidence\": 0.9819715619087219\n },\n {\n \"word\": \"難\",\n \"start\": 18.43221875,\n \"end\": 18.65221875,\n \"confidence\": 0.9143779277801514\n },\n {\n \"word\": \"し\",\n \"start\": 18.65221875,\n \"end\": 18.93221875,\n \"confidence\": 0.9932558536529541\n },\n {\n \"word\": \"かった\",\n \"start\": 18.93221875,\n \"end\": 19.232218749999998,\n \"confidence\": 0.9475598335266113\n },\n {\n \"word\": \"感\",\n \"start\": 19.232218749999998,\n \"end\": 19.81221875,\n \"confidence\": 0.7528156042098999\n },\n {\n \"word\": \"情\",\n \"start\": 19.81221875,\n \"end\": 20.13221875,\n \"confidence\": 0.9957336783409119\n },\n {\n \"word\": \"や\",\n \"start\": 20.13221875,\n \"end\": 20.31221875,\n \"confidence\": 0.9539394974708557\n },\n {\n \"word\": \"ニ\",\n \"start\": 20.31221875,\n \"end\": 20.47221875,\n \"confidence\": 0.9420691132545471\n },\n {\n \"word\": \"ュ\",\n \"start\": 20.47221875,\n \"end\": 20.53221875,\n \"confidence\": 0.9969981908798218\n },\n {\n \"word\": \"ア\",\n \"start\": 20.53221875,\n \"end\": 20.63221875,\n \"confidence\": 0.6907036304473877\n },\n {\n \"word\": \"ン\",\n \"start\": 20.63221875,\n \"end\": 20.69221875,\n \"confidence\": 0.99290531873703\n },\n {\n \"word\": \"ス\",\n \"start\": 20.69221875,\n \"end\": 20.79221875,\n \"confidence\": 0.9979546070098877\n },\n {\n \"word\": \"を\",\n \"start\": 20.79221875,\n \"end\": 20.892218749999998,\n \"confidence\": 0.9615700244903564\n },\n {\n \"word\": \"表\",\n \"start\": 20.892218749999998,\n \"end\": 21.072218749999998,\n \"confidence\": 0.9784479737281799\n },\n {\n \"word\": \"現\",\n \"start\": 21.072218749999998,\n \"end\": 21.31221875,\n \"confidence\": 0.996801495552063\n },\n {\n \"word\": \"し\",\n \"start\": 21.31221875,\n \"end\": 21.47221875,\n \"confidence\": 0.9380661845207214\n }\n ],\n \"model_info\": {\n \"diarization_model\": \"pyannote/speaker-diarization-3.1\",\n \"transcription_model\": \"faster-whisper-small\",\n \"translation_model\": \"google_translate\"\n }\n },\n {\n \"start_time\": 22.06971875,\n \"end_time\": 24.44909375,\n \"speaker_id\": \"SPEAKER_00\",\n \"original_text\": \"ユーザーの興味と理解を深めます。\",\n \"original_language\": \"ja\",\n \"translated_text\": \"It will enhance the user's interest and understanding.\",\n \"confidence_diarization\": 1.0,\n \"confidence_transcription\": -0.21058611944317818,\n \"confidence_translation\": 0.8,\n \"word_timestamps\": [\n {\n \"word\": \"ユ\",\n \"start\": 22.06971875,\n \"end\": 22.32971875,\n \"confidence\": 0.9343394935131073\n },\n {\n \"word\": \"ー\",\n \"start\": 22.32971875,\n \"end\": 22.36971875,\n \"confidence\": 0.9572596549987793\n },\n {\n \"word\": \"ザ\",\n \"start\": 22.36971875,\n \"end\": 22.46971875,\n \"confidence\": 0.9946682453155518\n },\n {\n \"word\": \"ー\",\n \"start\": 22.46971875,\n \"end\": 22.56971875,\n \"confidence\": 0.9885249733924866\n },\n {\n \"word\": \"の\",\n \"start\": 22.56971875,\n \"end\": 22.68971875,\n \"confidence\": 0.9828354716300964\n },\n {\n \"word\": \"興\",\n \"start\": 22.68971875,\n \"end\": 23.04971875,\n \"confidence\": 0.9197956323623657\n },\n {\n \"word\": \"味\",\n \"start\": 23.04971875,\n \"end\": 23.26971875,\n \"confidence\": 0.9995653033256531\n },\n {\n \"word\": \"と\",\n \"start\": 23.26971875,\n \"end\": 23.40971875,\n \"confidence\": 0.9928146600723267\n },\n {\n \"word\": \"理\",\n \"start\": 23.40971875,\n \"end\": 23.54971875,\n \"confidence\": 0.984175980091095\n },\n {\n \"word\": \"解\",\n \"start\": 23.54971875,\n \"end\": 23.76971875,\n \"confidence\": 0.999264657497406\n },\n {\n \"word\": \"を\",\n \"start\": 23.76971875,\n \"end\": 23.90971875,\n \"confidence\": 0.9952150583267212\n },\n {\n \"word\": \"深\",\n \"start\": 23.90971875,\n \"end\": 24.02971875,\n \"confidence\": 0.9548993110656738\n },\n {\n \"word\": \"め\",\n \"start\": 24.02971875,\n \"end\": 24.22971875,\n \"confidence\": 0.9892219305038452\n },\n {\n \"word\": \"ます。\",\n \"start\": 24.22971875,\n \"end\": 24.38971875,\n \"confidence\": 0.9906104207038879\n }\n ],\n \"model_info\": {\n \"diarization_model\": \"pyannote/speaker-diarization-3.1\",\n \"transcription_model\": \"faster-whisper-small\",\n \"translation_model\": \"google_translate\"\n }\n },\n {\n \"start_time\": 25.47846875,\n \"end_time\": 25.832843750000002,\n \"speaker_id\": \"SPEAKER_00\",\n \"original_text\": \"見る\",\n \"original_language\": \"ja\",\n \"translated_text\": \"See.\",\n \"confidence_diarization\": 1.0,\n \"confidence_transcription\": -0.4798548221588135,\n \"confidence_translation\": 0.8,\n \"word_timestamps\": [\n {\n \"word\": \"見\",\n \"start\": 25.47846875,\n \"end\": 25.65846875,\n \"confidence\": 0.5454539060592651\n },\n {\n \"word\": \"る\",\n \"start\": 25.65846875,\n \"end\": 25.738468750000003,\n \"confidence\": 0.9957653284072876\n }\n ],\n \"model_info\": {\n \"diarization_model\": \"pyannote/speaker-diarization-3.1\",\n \"transcription_model\": \"faster-whisper-small\",\n \"translation_model\": \"google_translate\"\n }\n },\n {\n \"start_time\": 26.204093750000002,\n \"end_time\": 26.65971875,\n \"speaker_id\": \"SPEAKER_00\",\n \"original_text\": \"聞く\",\n \"original_language\": \"ja\",\n \"translated_text\": \"Listen.\",\n \"confidence_diarization\": 1.0,\n \"confidence_transcription\": -0.47348871231079104,\n \"confidence_translation\": 0.8,\n \"word_timestamps\": [\n {\n \"word\": \"聞\",\n \"start\": 26.204093750000002,\n \"end\": 26.38409375,\n \"confidence\": 0.3832226097583771\n },\n {\n \"word\": \"く\",\n \"start\": 26.38409375,\n \"end\": 26.524093750000002,\n \"confidence\": 0.9974996447563171\n }\n ],\n \"model_info\": {\n \"diarization_model\": \"pyannote/speaker-diarization-3.1\",\n \"transcription_model\": \"faster-whisper-small\",\n \"translation_model\": \"google_translate\"\n }\n },\n {\n \"start_time\": 26.96346875,\n \"end_time\": 28.617218750000003,\n \"speaker_id\": \"SPEAKER_00\",\n \"original_text\": \"理解するウェブサイトへ\",\n \"original_language\": \"ja\",\n \"translated_text\": \"To a website that understands.\",\n \"confidence_diarization\": 1.0,\n \"confidence_transcription\": -0.27092968500577486,\n \"confidence_translation\": 0.8,\n \"word_timestamps\": [\n {\n \"word\": \"理\",\n \"start\": 26.96346875,\n \"end\": 27.14346875,\n \"confidence\": 0.4825628995895386\n },\n {\n \"word\": \"解\",\n \"start\": 27.14346875,\n \"end\": 27.36346875,\n \"confidence\": 0.9988553524017334\n },\n {\n \"word\": \"する\",\n \"start\": 27.36346875,\n \"end\": 27.64346875,\n \"confidence\": 0.9615910649299622\n },\n {\n \"word\": \"ウ\",\n \"start\": 27.64346875,\n \"end\": 27.903468750000002,\n \"confidence\": 0.4475053548812866\n },\n {\n \"word\": \"ェ\",\n \"start\": 27.903468750000002,\n \"end\": 28.00346875,\n \"confidence\": 0.9590348601341248\n },\n {\n \"word\": \"ブ\",\n \"start\": 28.00346875,\n \"end\": 28.08346875,\n \"confidence\": 0.989797830581665\n },\n {\n \"word\": \"サ\",\n \"start\": 28.08346875,\n \"end\": 28.28346875,\n \"confidence\": 0.9823185205459595\n },\n {\n \"word\": \"イ\",\n \"start\": 28.28346875,\n \"end\": 28.34346875,\n \"confidence\": 0.998434841632843\n },\n {\n \"word\": \"ト\",\n \"start\": 28.34346875,\n \"end\": 28.48346875,\n \"confidence\": 0.9974147081375122\n },\n {\n \"word\": \"へ\",\n \"start\": 28.48346875,\n \"end\": 28.58346875,\n \"confidence\": 0.9876385927200317\n }\n ],\n \"model_info\": {\n \"diarization_model\": \"pyannote/speaker-diarization-3.1\",\n \"transcription_model\": \"faster-whisper-small\",\n \"translation_model\": \"google_translate\"\n }\n },\n {\n \"start_time\": 29.24159375,\n \"end_time\": 31.90784375,\n \"speaker_id\": \"SPEAKER_00\",\n \"original_text\": \"音声メッセージが人の心を動かします\",\n \"original_language\": \"ja\",\n \"translated_text\": \"And that's what I'm talking about.\",\n \"confidence_diarization\": 1.0,\n \"confidence_transcription\": -0.23565174551571116,\n \"confidence_translation\": 0.8,\n \"word_timestamps\": [\n {\n \"word\": \"音\",\n \"start\": 29.24159375,\n \"end\": 29.42159375,\n \"confidence\": 0.9116391539573669\n },\n {\n \"word\": \"声\",\n \"start\": 29.42159375,\n \"end\": 29.64159375,\n \"confidence\": 0.979734480381012\n },\n {\n \"word\": \"メ\",\n \"start\": 29.64159375,\n \"end\": 29.78159375,\n \"confidence\": 0.896361768245697\n },\n {\n \"word\": \"ッ\",\n \"start\": 29.78159375,\n \"end\": 29.86159375,\n \"confidence\": 0.9995806813240051\n },\n {\n \"word\": \"セ\",\n \"start\": 29.86159375,\n \"end\": 29.96159375,\n \"confidence\": 0.9946938157081604\n },\n {\n \"word\": \"ージ\",\n \"start\": 29.96159375,\n \"end\": 30.08159375,\n \"confidence\": 0.9994053840637207\n },\n {\n \"word\": \"が\",\n \"start\": 30.08159375,\n \"end\": 30.28159375,\n \"confidence\": 0.9612740278244019\n },\n {\n \"word\": \"人\",\n \"start\": 30.28159375,\n \"end\": 30.56159375,\n \"confidence\": 0.839630663394928\n },\n {\n \"word\": \"の\",\n \"start\": 30.56159375,\n \"end\": 30.78159375,\n \"confidence\": 0.9984166622161865\n },\n {\n \"word\": \"心\",\n \"start\": 30.78159375,\n \"end\": 31.00159375,\n \"confidence\": 0.9308077692985535\n },\n {\n \"word\": \"を\",\n \"start\": 31.00159375,\n \"end\": 31.28159375,\n \"confidence\": 0.9952632188796997\n },\n {\n \"word\": \"動\",\n \"start\": 31.28159375,\n \"end\": 31.42159375,\n \"confidence\": 0.9899610280990601\n },\n {\n \"word\": \"か\",\n \"start\": 31.42159375,\n \"end\": 31.58159375,\n \"confidence\": 0.9986295700073242\n },\n {\n \"word\": \"します\",\n \"start\": 31.58159375,\n \"end\": 31.74159375,\n \"confidence\": 0.9892330169677734\n }\n ],\n \"model_info\": {\n \"diarization_model\": \"pyannote/speaker-diarization-3.1\",\n \"transcription_model\": \"faster-whisper-small\",\n \"translation_model\": \"google_translate\"\n }\n }\n ],\n \"speakers\": {\n \"SPEAKER_00\": {\n \"total_speaking_time\": 26.021250000000002,\n \"number_of_turns\": 12,\n \"longest_turn\": 5.248125000000002,\n \"shortest_turn\": 0.354375000000001,\n \"languages\": [\n \"ja\"\n ],\n \"average_turn_duration\": 2.1684375\n }\n },\n \"languages\": {\n \"ja\": {\n \"speaking_time\": 26.021250000000002,\n \"segment_count\": 12,\n \"speakers\": [\n \"SPEAKER_00\"\n ]\n }\n }\n}", "srt_original": "1\n00:00:00,402 --> 00:00:04,772\n[JA] 音声メッセージが既存のウェブサイトを超えたコミュニケーションを実現。\n\n2\n00:00:05,515 --> 00:00:07,388\n[JA] 目で見るだけだったウェブサイトに\n\n3\n00:00:07,624 --> 00:00:09,852\n[JA] 音声情報をインクルードすることで\n\n4\n00:00:10,274 --> 00:00:12,315\n[JA] 情報に新しい価値を与え\n\n5\n00:00:12,366 --> 00:00:14,729\n[JA] 他者との差別化に効果を発揮します\n\n6\n00:00:15,674 --> 00:00:16,062\n[JA] また!\n\n7\n00:00:16,332 --> 00:00:21,580\n[JA] 文字やグラフィックだけでは伝えることの難しかった感情やニュアンスを表現し\n\n8\n00:00:22,069 --> 00:00:24,449\n[JA] ユーザーの興味と理解を深めます。\n\n9\n00:00:25,478 --> 00:00:25,832\n[JA] 見る\n\n10\n00:00:26,204 --> 00:00:26,659\n[JA] 聞く\n\n11\n00:00:26,963 --> 00:00:28,617\n[JA] 理解するウェブサイトへ\n\n12\n00:00:29,241 --> 00:00:31,907\n[JA] 音声メッセージが人の心を動かします\n", "srt_translated": "1\n00:00:00,402 --> 00:00:04,772\nThe audio message will bring out communication beyond the existing website.\n\n2\n00:00:05,515 --> 00:00:07,388\nI'm going to show you what I'm doing.\n\n3\n00:00:07,624 --> 00:00:09,852\nWe're going to be able to do that in the next video.\n\n4\n00:00:10,274 --> 00:00:12,315\nAnd that's what we're going to do.\n\n5\n00:00:12,366 --> 00:00:14,729\nIt's not just about being different from other people.\n\n6\n00:00:15,674 --> 00:00:16,062\nAgain!\n\n7\n00:00:16,332 --> 00:00:21,580\nIt's not just writing, it's graphic.\n\n8\n00:00:22,069 --> 00:00:24,449\nIt will enhance the user's interest and understanding.\n\n9\n00:00:25,478 --> 00:00:25,832\nSee.\n\n10\n00:00:26,204 --> 00:00:26,659\nListen.\n\n11\n00:00:26,963 --> 00:00:28,617\nTo a website that understands.\n\n12\n00:00:29,241 --> 00:00:31,907\nAnd that's what I'm talking about.\n", "text": "================================================================================\nMULTILINGUAL AUDIO INTELLIGENCE ANALYSIS\n================================================================================\n\nAudio File: Yuri_Kizaki.mp3\nAnalysis Date: 2025-09-02T16:18:58.085380\nDuration: 32.4s\nSample Rate: 44100 Hz\nChannels: 1\n\nANALYSIS SUMMARY\n----------------------------------------\nTotal Speakers: 1\nLanguages Detected: ja\nTotal Segments: 12\nSpeech Duration: 26.0s\nSpeech Ratio: 81.6%\nProcessing Time: Unknown\n\nSPEAKER BREAKDOWN\n----------------------------------------\nSpeaker 00:\n Speaking Time: 26.0s\n Number of Turns: 12\n Average Turn: 2.2s\n Longest Turn: 5.2s\n Languages: ja\n\nFULL TRANSCRIPT\n================================================================================\n\n# 1 [0.4s - 4.8s] Speaker 00\n Original (ja): 音声メッセージが既存のウェブサイトを超えたコミュニケーションを実現。\n Translation: The audio message will bring out communication beyond the existing website.\n Confidence: D:1.00 T:-0.18 TR:0.80\n\n# 2 [5.5s - 7.4s] Speaker 00\n Original (ja): 目で見るだけだったウェブサイトに\n Translation: I'm going to show you what I'm doing.\n Confidence: D:1.00 T:-0.22 TR:0.80\n\n# 3 [7.6s - 9.9s] Speaker 00\n Original (ja): 音声情報をインクルードすることで\n Translation: We're going to be able to do that in the next video.\n Confidence: D:1.00 T:-0.24 TR:0.80\n\n# 4 [10.3s - 12.3s] Speaker 00\n Original (ja): 情報に新しい価値を与え\n Translation: And that's what we're going to do.\n Confidence: D:1.00 T:-0.12 TR:0.80\n\n# 5 [12.4s - 14.7s] Speaker 00\n Original (ja): 他者との差別化に効果を発揮します\n Translation: It's not just about being different from other people.\n Confidence: D:1.00 T:-0.23 TR:0.80\n\n# 6 [15.7s - 16.1s] Speaker 00\n Original (ja): また!\n Translation: Again!\n Confidence: D:1.00 T:-0.48 TR:0.80\n\n# 7 [16.3s - 21.6s] Speaker 00\n Original (ja): 文字やグラフィックだけでは伝えることの難しかった感情やニュアンスを表現し\n Translation: It's not just writing, it's graphic.\n Confidence: D:1.00 T:-0.16 TR:0.80\n\n# 8 [22.1s - 24.4s] Speaker 00\n Original (ja): ユーザーの興味と理解を深めます。\n Translation: It will enhance the user's interest and understanding.\n Confidence: D:1.00 T:-0.21 TR:0.80\n\n# 9 [25.5s - 25.8s] Speaker 00\n Original (ja): 見る\n Translation: See.\n Confidence: D:1.00 T:-0.48 TR:0.80\n\n# 10 [26.2s - 26.7s] Speaker 00\n Original (ja): 聞く\n Translation: Listen.\n Confidence: D:1.00 T:-0.47 TR:0.80\n\n# 11 [27.0s - 28.6s] Speaker 00\n Original (ja): 理解するウェブサイトへ\n Translation: To a website that understands.\n Confidence: D:1.00 T:-0.27 TR:0.80\n\n# 12 [29.2s - 31.9s] Speaker 00\n Original (ja): 音声メッセージが人の心を動かします\n Translation: And that's what I'm talking about.\n Confidence: D:1.00 T:-0.24 TR:0.80\n\n================================================================================\nGenerated by Multilingual Audio Intelligence System\n================================================================================", "csv": "segment_id,start_time,end_time,duration,speaker_id,original_language,original_text,translated_text,confidence_diarization,confidence_transcription,confidence_translation,word_count_original,word_count_translated\r\n1,0.40221875,4.77284375,4.3706249999999995,SPEAKER_00,ja,音声メッセージが既存のウェブサイトを超えたコミュニケーションを実現。,The audio message will bring out communication beyond the existing website.,1.0,-0.1825541319946448,0.8,1,11\r\n2,5.5153437499999995,7.388468750000001,1.8731250000000017,SPEAKER_00,ja,目で見るだけだったウェブサイトに,I'm going to show you what I'm doing.,1.0,-0.22203674035913804,0.8,1,8\r\n3,7.624718750000001,9.852218750000002,2.227500000000001,SPEAKER_00,ja,音声情報をインクルードすることで,We're going to be able to do that in the next video.,1.0,-0.2369275689125061,0.8,1,12\r\n4,10.274093750000002,12.31596875,2.0418749999999974,SPEAKER_00,ja,情報に新しい価値を与え,And that's what we're going to do.,1.0,-0.11563345324248075,0.8,1,7\r\n5,12.36659375,14.72909375,2.3625000000000007,SPEAKER_00,ja,他者との差別化に効果を発揮します,It's not just about being different from other people.,1.0,-0.2329371053921549,0.8,1,9\r\n6,15.67409375,16.06221875,0.3881249999999987,SPEAKER_00,ja,また!,Again!,1.0,-0.4752265453338623,0.8,1,1\r\n7,16.33221875,21.58034375,5.248125000000002,SPEAKER_00,ja,文字やグラフィックだけでは伝えることの難しかった感情やニュアンスを表現し,\"It's not just writing, it's graphic.\",1.0,-0.16042621207959723,0.8,1,6\r\n8,22.06971875,24.44909375,2.3793749999999996,SPEAKER_00,ja,ユーザーの興味と理解を深めます。,It will enhance the user's interest and understanding.,1.0,-0.21058611944317818,0.8,1,8\r\n9,25.47846875,25.832843750000002,0.354375000000001,SPEAKER_00,ja,見る,See.,1.0,-0.4798548221588135,0.8,1,1\r\n10,26.204093750000002,26.65971875,0.4556249999999977,SPEAKER_00,ja,聞く,Listen.,1.0,-0.47348871231079104,0.8,1,1\r\n11,26.96346875,28.617218750000003,1.6537500000000023,SPEAKER_00,ja,理解するウェブサイトへ,To a website that understands.,1.0,-0.27092968500577486,0.8,1,5\r\n12,29.24159375,31.90784375,2.6662500000000016,SPEAKER_00,ja,音声メッセージが人の心を動かします,And that's what I'm talking about.,1.0,-0.23565174551571116,0.8,1,6\r\n", "timeline": "{\n \"title\": {\n \"text\": {\n \"headline\": \"Audio Analysis: Yuri_Kizaki.mp3\",\n \"text\": \"Interactive timeline of speaker segments and transcription\"\n }\n },\n \"events\": [\n {\n \"start_date\": {\n \"second\": 0\n },\n \"end_date\": {\n \"second\": 4\n },\n \"text\": {\n \"headline\": \"Speaker 00 (ja)\",\n \"text\": \"

Original: 音声メッセージが既存のウェブサイトを超えたコミュニケーションを実現。

Translation: The audio message will bring out communication beyond the existing website.

Duration: 4.4s, Confidence: -0.18

\"\n },\n \"group\": \"SPEAKER_00\",\n \"media\": {\n \"caption\": \"Segment 1: 0.4s - 4.8s\"\n }\n },\n {\n \"start_date\": {\n \"second\": 5\n },\n \"end_date\": {\n \"second\": 7\n },\n \"text\": {\n \"headline\": \"Speaker 00 (ja)\",\n \"text\": \"

Original: 目で見るだけだったウェブサイトに

Translation: I'm going to show you what I'm doing.

Duration: 1.9s, Confidence: -0.22

\"\n },\n \"group\": \"SPEAKER_00\",\n \"media\": {\n \"caption\": \"Segment 2: 5.5s - 7.4s\"\n }\n },\n {\n \"start_date\": {\n \"second\": 7\n },\n \"end_date\": {\n \"second\": 9\n },\n \"text\": {\n \"headline\": \"Speaker 00 (ja)\",\n \"text\": \"

Original: 音声情報をインクルードすることで

Translation: We're going to be able to do that in the next video.

Duration: 2.2s, Confidence: -0.24

\"\n },\n \"group\": \"SPEAKER_00\",\n \"media\": {\n \"caption\": \"Segment 3: 7.6s - 9.9s\"\n }\n },\n {\n \"start_date\": {\n \"second\": 10\n },\n \"end_date\": {\n \"second\": 12\n },\n \"text\": {\n \"headline\": \"Speaker 00 (ja)\",\n \"text\": \"

Original: 情報に新しい価値を与え

Translation: And that's what we're going to do.

Duration: 2.0s, Confidence: -0.12

\"\n },\n \"group\": \"SPEAKER_00\",\n \"media\": {\n \"caption\": \"Segment 4: 10.3s - 12.3s\"\n }\n },\n {\n \"start_date\": {\n \"second\": 12\n },\n \"end_date\": {\n \"second\": 14\n },\n \"text\": {\n \"headline\": \"Speaker 00 (ja)\",\n \"text\": \"

Original: 他者との差別化に効果を発揮します

Translation: It's not just about being different from other people.

Duration: 2.4s, Confidence: -0.23

\"\n },\n \"group\": \"SPEAKER_00\",\n \"media\": {\n \"caption\": \"Segment 5: 12.4s - 14.7s\"\n }\n },\n {\n \"start_date\": {\n \"second\": 15\n },\n \"end_date\": {\n \"second\": 16\n },\n \"text\": {\n \"headline\": \"Speaker 00 (ja)\",\n \"text\": \"

Original: また!

Translation: Again!

Duration: 0.4s, Confidence: -0.48

\"\n },\n \"group\": \"SPEAKER_00\",\n \"media\": {\n \"caption\": \"Segment 6: 15.7s - 16.1s\"\n }\n },\n {\n \"start_date\": {\n \"second\": 16\n },\n \"end_date\": {\n \"second\": 21\n },\n \"text\": {\n \"headline\": \"Speaker 00 (ja)\",\n \"text\": \"

Original: 文字やグラフィックだけでは伝えることの難しかった感情やニュアンスを表現し

Translation: It's not just writing, it's graphic.

Duration: 5.2s, Confidence: -0.16

\"\n },\n \"group\": \"SPEAKER_00\",\n \"media\": {\n \"caption\": \"Segment 7: 16.3s - 21.6s\"\n }\n },\n {\n \"start_date\": {\n \"second\": 22\n },\n \"end_date\": {\n \"second\": 24\n },\n \"text\": {\n \"headline\": \"Speaker 00 (ja)\",\n \"text\": \"

Original: ユーザーの興味と理解を深めます。

Translation: It will enhance the user's interest and understanding.

Duration: 2.4s, Confidence: -0.21

\"\n },\n \"group\": \"SPEAKER_00\",\n \"media\": {\n \"caption\": \"Segment 8: 22.1s - 24.4s\"\n }\n },\n {\n \"start_date\": {\n \"second\": 25\n },\n \"end_date\": {\n \"second\": 25\n },\n \"text\": {\n \"headline\": \"Speaker 00 (ja)\",\n \"text\": \"

Original: 見る

Translation: See.

Duration: 0.4s, Confidence: -0.48

\"\n },\n \"group\": \"SPEAKER_00\",\n \"media\": {\n \"caption\": \"Segment 9: 25.5s - 25.8s\"\n }\n },\n {\n \"start_date\": {\n \"second\": 26\n },\n \"end_date\": {\n \"second\": 26\n },\n \"text\": {\n \"headline\": \"Speaker 00 (ja)\",\n \"text\": \"

Original: 聞く

Translation: Listen.

Duration: 0.5s, Confidence: -0.47

\"\n },\n \"group\": \"SPEAKER_00\",\n \"media\": {\n \"caption\": \"Segment 10: 26.2s - 26.7s\"\n }\n },\n {\n \"start_date\": {\n \"second\": 26\n },\n \"end_date\": {\n \"second\": 28\n },\n \"text\": {\n \"headline\": \"Speaker 00 (ja)\",\n \"text\": \"

Original: 理解するウェブサイトへ

Translation: To a website that understands.

Duration: 1.7s, Confidence: -0.27

\"\n },\n \"group\": \"SPEAKER_00\",\n \"media\": {\n \"caption\": \"Segment 11: 27.0s - 28.6s\"\n }\n },\n {\n \"start_date\": {\n \"second\": 29\n },\n \"end_date\": {\n \"second\": 31\n },\n \"text\": {\n \"headline\": \"Speaker 00 (ja)\",\n \"text\": \"

Original: 音声メッセージが人の心を動かします

Translation: And that's what I'm talking about.

Duration: 2.7s, Confidence: -0.24

\"\n },\n \"group\": \"SPEAKER_00\",\n \"media\": {\n \"caption\": \"Segment 12: 29.2s - 31.9s\"\n }\n }\n ]\n}", "summary": "ANALYSIS SUMMARY FOR Yuri_Kizaki.mp3\n==================================================\n\n• 1 speakers detected\n• 12 speech segments identified\n• 1 languages detected: ja\n• 81.6% of audio contains speech\n\nSPEAKER BREAKDOWN:\n• Speaker 00: 26.0s (100.0%) across 12 turns\n\nKEY INSIGHTS:\n• Most active speaker: Speaker 00\n• Longest speaking turn: 5.2s by Speaker 00\n• Average transcription confidence: -0.27" }, "saved_files": { "json": "results\\Yuri_Kizaki.json", "text": "results\\Yuri_Kizaki.txt", "summary": "results\\Yuri_Kizaki.summary.txt" }, "processed_segments": [ "ProcessedSegment(start_time=0.40221875, end_time=4.77284375, speaker_id='SPEAKER_00', original_text='音声メッセージが既存のウェブサイトを超えたコミュニケーションを実現。', original_language='ja', translated_text='The audio message will bring out communication beyond the existing website.', confidence_diarization=1.0, confidence_transcription=-0.1825541319946448, confidence_translation=0.8, word_timestamps=[{'word': '音', 'start': 0.40221875, 'end': 0.56221875, 'confidence': 0.8530172109603882}, {'word': '声', 'start': 0.56221875, 'end': 0.80221875, 'confidence': 0.9917272329330444}, {'word': 'メ', 'start': 0.80221875, 'end': 0.9422187500000001, 'confidence': 0.9574464559555054}, {'word': 'ッ', 'start': 0.9422187500000001, 'end': 1.02221875, 'confidence': 0.999119222164154}, {'word': 'セ', 'start': 1.02221875, 'end': 1.14221875, 'confidence': 0.99460768699646}, {'word': 'ージ', 'start': 1.14221875, 'end': 1.30221875, 'confidence': 0.9997381567955017}, {'word': 'が', 'start': 1.30221875, 'end': 1.5222187500000002, 'confidence': 0.9662947654724121}, {'word': '既', 'start': 1.5222187500000002, 'end': 1.92221875, 'confidence': 0.7296531945466995}, {'word': '存', 'start': 1.92221875, 'end': 2.08221875, 'confidence': 0.9589823484420776}, {'word': 'の', 'start': 2.08221875, 'end': 2.20221875, 'confidence': 0.9912187457084656}, {'word': 'ウ', 'start': 2.20221875, 'end': 2.3022187499999998, 'confidence': 0.6959699988365173}, {'word': 'ェ', 'start': 2.3022187499999998, 'end': 2.36221875, 'confidence': 0.9874258041381836}, {'word': 'ブ', 'start': 2.36221875, 'end': 2.48221875, 'confidence': 0.9893200397491455}, {'word': 'サ', 'start': 2.48221875, 'end': 2.64221875, 'confidence': 0.9838968515396118}, {'word': 'イ', 'start': 2.64221875, 'end': 2.7222187499999997, 'confidence': 0.9970263838768005}, {'word': 'ト', 'start': 2.7222187499999997, 'end': 2.86221875, 'confidence': 0.9971777200698853}, {'word': 'を', 'start': 2.86221875, 'end': 2.94221875, 'confidence': 0.9877551198005676}, {'word': '超', 'start': 2.94221875, 'end': 3.04221875, 'confidence': 0.6848042011260986}, {'word': 'え', 'start': 3.04221875, 'end': 3.1822187499999997, 'confidence': 0.9907885193824768}, {'word': 'た', 'start': 3.1822187499999997, 'end': 3.2822187499999997, 'confidence': 0.9983263611793518}, {'word': 'コ', 'start': 3.2822187499999997, 'end': 3.44221875, 'confidence': 0.9066019058227539}, {'word': 'ミ', 'start': 3.44221875, 'end': 3.54221875, 'confidence': 0.9985296726226807}, {'word': 'ュ', 'start': 3.54221875, 'end': 3.58221875, 'confidence': 0.9981721639633179}, {'word': 'ニ', 'start': 3.58221875, 'end': 3.6622187499999996, 'confidence': 0.9988634586334229}, {'word': 'ケ', 'start': 3.6622187499999996, 'end': 3.8222187499999998, 'confidence': 0.9971752166748047}, {'word': 'ー', 'start': 3.8222187499999998, 'end': 3.90221875, 'confidence': 0.9970790147781372}, {'word': 'ショ', 'start': 3.90221875, 'end': 4.00221875, 'confidence': 0.9993009567260742}, {'word': 'ン', 'start': 4.00221875, 'end': 4.1022187500000005, 'confidence': 0.9991468191146851}, {'word': 'を', 'start': 4.1022187500000005, 'end': 4.18221875, 'confidence': 0.991553008556366}, {'word': '実', 'start': 4.18221875, 'end': 4.36221875, 'confidence': 0.9924994111061096}, {'word': '現。', 'start': 4.36221875, 'end': 4.6022187500000005, 'confidence': 0.9942215085029602}], model_info={'diarization_model': 'pyannote/speaker-diarization-3.1', 'transcription_model': 'faster-whisper-small', 'translation_model': 'google_translate'})", "ProcessedSegment(start_time=5.5153437499999995, end_time=7.388468750000001, speaker_id='SPEAKER_00', original_text='目で見るだけだったウェブサイトに', original_language='ja', translated_text=\"I'm going to show you what I'm doing.\", confidence_diarization=1.0, confidence_transcription=-0.22203674035913804, confidence_translation=0.8, word_timestamps=[{'word': '目', 'start': 5.5153437499999995, 'end': 5.655343749999999, 'confidence': 0.8701557517051697}, {'word': 'で', 'start': 5.655343749999999, 'end': 5.815343749999999, 'confidence': 0.991607666015625}, {'word': '見', 'start': 5.815343749999999, 'end': 5.9353437499999995, 'confidence': 0.9280027151107788}, {'word': 'る', 'start': 5.9353437499999995, 'end': 6.05534375, 'confidence': 0.9964483976364136}, {'word': 'だけ', 'start': 6.05534375, 'end': 6.235343749999999, 'confidence': 0.9943233728408813}, {'word': 'だ', 'start': 6.235343749999999, 'end': 6.4353437499999995, 'confidence': 0.9976925849914551}, {'word': 'った', 'start': 6.4353437499999995, 'end': 6.57534375, 'confidence': 0.9989917874336243}, {'word': 'ウ', 'start': 6.57534375, 'end': 6.67534375, 'confidence': 0.4343600571155548}, {'word': 'ェ', 'start': 6.67534375, 'end': 6.735343749999999, 'confidence': 0.9842584133148193}, {'word': 'ブ', 'start': 6.735343749999999, 'end': 6.83534375, 'confidence': 0.9933525323867798}, {'word': 'サ', 'start': 6.83534375, 'end': 7.0153437499999995, 'confidence': 0.9906386137008667}, {'word': 'イ', 'start': 7.0153437499999995, 'end': 7.07534375, 'confidence': 0.9990501999855042}, {'word': 'ト', 'start': 7.07534375, 'end': 7.195343749999999, 'confidence': 0.9961349964141846}, {'word': 'に', 'start': 7.195343749999999, 'end': 7.315343749999999, 'confidence': 0.989922821521759}], model_info={'diarization_model': 'pyannote/speaker-diarization-3.1', 'transcription_model': 'faster-whisper-small', 'translation_model': 'google_translate'})", "ProcessedSegment(start_time=7.624718750000001, end_time=9.852218750000002, speaker_id='SPEAKER_00', original_text='音声情報をインクルードすることで', original_language='ja', translated_text=\"We're going to be able to do that in the next video.\", confidence_diarization=1.0, confidence_transcription=-0.2369275689125061, confidence_translation=0.8, word_timestamps=[{'word': '音', 'start': 7.624718750000001, 'end': 7.7847187500000015, 'confidence': 0.9499445557594299}, {'word': '声', 'start': 7.7847187500000015, 'end': 8.004718750000002, 'confidence': 0.9357801079750061}, {'word': '情', 'start': 8.004718750000002, 'end': 8.164718750000002, 'confidence': 0.9815613627433777}, {'word': '報', 'start': 8.164718750000002, 'end': 8.40471875, 'confidence': 0.9961434602737427}, {'word': 'を', 'start': 8.40471875, 'end': 8.544718750000001, 'confidence': 0.992678165435791}, {'word': 'イ', 'start': 8.544718750000001, 'end': 8.684718750000002, 'confidence': 0.9322373270988464}, {'word': 'ン', 'start': 8.684718750000002, 'end': 8.74471875, 'confidence': 0.9673494696617126}, {'word': 'ク', 'start': 8.74471875, 'end': 8.844718750000002, 'confidence': 0.9965403079986572}, {'word': 'ル', 'start': 8.844718750000002, 'end': 8.944718750000002, 'confidence': 0.9498746395111084}, {'word': 'ード', 'start': 8.944718750000002, 'end': 9.124718750000001, 'confidence': 0.9774163961410522}, {'word': 'する', 'start': 9.124718750000001, 'end': 9.364718750000002, 'confidence': 0.9932113885879517}, {'word': 'こと', 'start': 9.364718750000002, 'end': 9.56471875, 'confidence': 0.9621437191963196}, {'word': 'で', 'start': 9.56471875, 'end': 9.764718750000002, 'confidence': 0.9964655637741089}], model_info={'diarization_model': 'pyannote/speaker-diarization-3.1', 'transcription_model': 'faster-whisper-small', 'translation_model': 'google_translate'})", "ProcessedSegment(start_time=10.274093750000002, end_time=12.31596875, speaker_id='SPEAKER_00', original_text='情報に新しい価値を与え', original_language='ja', translated_text=\"And that's what we're going to do.\", confidence_diarization=1.0, confidence_transcription=-0.11563345324248075, confidence_translation=0.8, word_timestamps=[{'word': '情', 'start': 10.274093750000002, 'end': 10.474093750000002, 'confidence': 0.9788916110992432}, {'word': '報', 'start': 10.474093750000002, 'end': 10.694093750000002, 'confidence': 0.9990907907485962}, {'word': 'に', 'start': 10.694093750000002, 'end': 10.814093750000001, 'confidence': 0.9892839789390564}, {'word': '新', 'start': 10.814093750000001, 'end': 11.014093750000002, 'confidence': 0.9793343544006348}, {'word': 'しい', 'start': 11.014093750000002, 'end': 11.394093750000003, 'confidence': 0.9975306391716003}, {'word': '価', 'start': 11.394093750000003, 'end': 11.574093750000003, 'confidence': 0.981714278459549}, {'word': '値', 'start': 11.574093750000003, 'end': 11.754093750000003, 'confidence': 0.9989857375621796}, {'word': 'を', 'start': 11.754093750000003, 'end': 11.854093750000002, 'confidence': 0.9980254173278809}, {'word': '与', 'start': 11.854093750000002, 'end': 12.114093750000002, 'confidence': 0.9476390182971954}, {'word': 'え', 'start': 12.114093750000002, 'end': 12.194093750000002, 'confidence': 0.9922704696655273}], model_info={'diarization_model': 'pyannote/speaker-diarization-3.1', 'transcription_model': 'faster-whisper-small', 'translation_model': 'google_translate'})", "ProcessedSegment(start_time=12.36659375, end_time=14.72909375, speaker_id='SPEAKER_00', original_text='他者との差別化に効果を発揮します', original_language='ja', translated_text=\"It's not just about being different from other people.\", confidence_diarization=1.0, confidence_transcription=-0.2329371053921549, confidence_translation=0.8, word_timestamps=[{'word': '他', 'start': 12.36659375, 'end': 12.56659375, 'confidence': 0.7133576273918152}, {'word': '者', 'start': 12.56659375, 'end': 12.72659375, 'confidence': 0.594456672668457}, {'word': 'と', 'start': 12.72659375, 'end': 12.84659375, 'confidence': 0.9945782423019409}, {'word': 'の', 'start': 12.84659375, 'end': 12.96659375, 'confidence': 0.998796820640564}, {'word': '差', 'start': 12.96659375, 'end': 13.10659375, 'confidence': 0.9885448813438416}, {'word': '別', 'start': 13.10659375, 'end': 13.30659375, 'confidence': 0.9973207116127014}, {'word': '化', 'start': 13.30659375, 'end': 13.48659375, 'confidence': 0.9788604378700256}, {'word': 'に', 'start': 13.48659375, 'end': 13.60659375, 'confidence': 0.9965766072273254}, {'word': '効', 'start': 13.60659375, 'end': 13.86659375, 'confidence': 0.9582771062850952}, {'word': '果', 'start': 13.86659375, 'end': 14.02659375, 'confidence': 0.9983495473861694}, {'word': 'を', 'start': 14.02659375, 'end': 14.12659375, 'confidence': 0.9957448840141296}, {'word': '発', 'start': 14.12659375, 'end': 14.246593749999999, 'confidence': 0.9888325929641724}, {'word': '揮', 'start': 14.246593749999999, 'end': 14.36659375, 'confidence': 0.9894059002399445}, {'word': 'します', 'start': 14.36659375, 'end': 14.54659375, 'confidence': 0.9909846782684326}], model_info={'diarization_model': 'pyannote/speaker-diarization-3.1', 'transcription_model': 'faster-whisper-small', 'translation_model': 'google_translate'})", "ProcessedSegment(start_time=15.67409375, end_time=16.06221875, speaker_id='SPEAKER_00', original_text='また!', original_language='ja', translated_text='Again!', confidence_diarization=1.0, confidence_transcription=-0.4752265453338623, confidence_translation=0.8, word_timestamps=[{'word': 'また!', 'start': 15.67409375, 'end': 15.894093750000001, 'confidence': 0.9813592433929443}], model_info={'diarization_model': 'pyannote/speaker-diarization-3.1', 'transcription_model': 'faster-whisper-small', 'translation_model': 'google_translate'})", "ProcessedSegment(start_time=16.33221875, end_time=21.58034375, speaker_id='SPEAKER_00', original_text='文字やグラフィックだけでは伝えることの難しかった感情やニュアンスを表現し', original_language='ja', translated_text=\"It's not just writing, it's graphic.\", confidence_diarization=1.0, confidence_transcription=-0.16042621207959723, confidence_translation=0.8, word_timestamps=[{'word': '文', 'start': 16.33221875, 'end': 16.53221875, 'confidence': 0.8754217624664307}, {'word': '字', 'start': 16.53221875, 'end': 16.69221875, 'confidence': 0.9960361123085022}, {'word': 'や', 'start': 16.69221875, 'end': 16.79221875, 'confidence': 0.9906545281410217}, {'word': 'グ', 'start': 16.79221875, 'end': 16.892218749999998, 'confidence': 0.9925161004066467}, {'word': 'ラ', 'start': 16.892218749999998, 'end': 17.01221875, 'confidence': 0.9981822967529297}, {'word': 'フ', 'start': 17.01221875, 'end': 17.072218749999998, 'confidence': 0.9955530762672424}, {'word': 'ィ', 'start': 17.072218749999998, 'end': 17.15221875, 'confidence': 0.9970651268959045}, {'word': 'ック', 'start': 17.15221875, 'end': 17.27221875, 'confidence': 0.9935983419418335}, {'word': 'だけ', 'start': 17.27221875, 'end': 17.45221875, 'confidence': 0.9928644895553589}, {'word': 'では', 'start': 17.45221875, 'end': 17.67221875, 'confidence': 0.9097373485565186}, {'word': '伝', 'start': 17.67221875, 'end': 17.91221875, 'confidence': 0.9866331815719604}, {'word': 'える', 'start': 17.91221875, 'end': 18.09221875, 'confidence': 0.9961875081062317}, {'word': 'こと', 'start': 18.09221875, 'end': 18.232218749999998, 'confidence': 0.8297985792160034}, {'word': 'の', 'start': 18.232218749999998, 'end': 18.43221875, 'confidence': 0.9819715619087219}, {'word': '難', 'start': 18.43221875, 'end': 18.65221875, 'confidence': 0.9143779277801514}, {'word': 'し', 'start': 18.65221875, 'end': 18.93221875, 'confidence': 0.9932558536529541}, {'word': 'かった', 'start': 18.93221875, 'end': 19.232218749999998, 'confidence': 0.9475598335266113}, {'word': '感', 'start': 19.232218749999998, 'end': 19.81221875, 'confidence': 0.7528156042098999}, {'word': '情', 'start': 19.81221875, 'end': 20.13221875, 'confidence': 0.9957336783409119}, {'word': 'や', 'start': 20.13221875, 'end': 20.31221875, 'confidence': 0.9539394974708557}, {'word': 'ニ', 'start': 20.31221875, 'end': 20.47221875, 'confidence': 0.9420691132545471}, {'word': 'ュ', 'start': 20.47221875, 'end': 20.53221875, 'confidence': 0.9969981908798218}, {'word': 'ア', 'start': 20.53221875, 'end': 20.63221875, 'confidence': 0.6907036304473877}, {'word': 'ン', 'start': 20.63221875, 'end': 20.69221875, 'confidence': 0.99290531873703}, {'word': 'ス', 'start': 20.69221875, 'end': 20.79221875, 'confidence': 0.9979546070098877}, {'word': 'を', 'start': 20.79221875, 'end': 20.892218749999998, 'confidence': 0.9615700244903564}, {'word': '表', 'start': 20.892218749999998, 'end': 21.072218749999998, 'confidence': 0.9784479737281799}, {'word': '現', 'start': 21.072218749999998, 'end': 21.31221875, 'confidence': 0.996801495552063}, {'word': 'し', 'start': 21.31221875, 'end': 21.47221875, 'confidence': 0.9380661845207214}], model_info={'diarization_model': 'pyannote/speaker-diarization-3.1', 'transcription_model': 'faster-whisper-small', 'translation_model': 'google_translate'})", "ProcessedSegment(start_time=22.06971875, end_time=24.44909375, speaker_id='SPEAKER_00', original_text='ユーザーの興味と理解を深めます。', original_language='ja', translated_text=\"It will enhance the user's interest and understanding.\", confidence_diarization=1.0, confidence_transcription=-0.21058611944317818, confidence_translation=0.8, word_timestamps=[{'word': 'ユ', 'start': 22.06971875, 'end': 22.32971875, 'confidence': 0.9343394935131073}, {'word': 'ー', 'start': 22.32971875, 'end': 22.36971875, 'confidence': 0.9572596549987793}, {'word': 'ザ', 'start': 22.36971875, 'end': 22.46971875, 'confidence': 0.9946682453155518}, {'word': 'ー', 'start': 22.46971875, 'end': 22.56971875, 'confidence': 0.9885249733924866}, {'word': 'の', 'start': 22.56971875, 'end': 22.68971875, 'confidence': 0.9828354716300964}, {'word': '興', 'start': 22.68971875, 'end': 23.04971875, 'confidence': 0.9197956323623657}, {'word': '味', 'start': 23.04971875, 'end': 23.26971875, 'confidence': 0.9995653033256531}, {'word': 'と', 'start': 23.26971875, 'end': 23.40971875, 'confidence': 0.9928146600723267}, {'word': '理', 'start': 23.40971875, 'end': 23.54971875, 'confidence': 0.984175980091095}, {'word': '解', 'start': 23.54971875, 'end': 23.76971875, 'confidence': 0.999264657497406}, {'word': 'を', 'start': 23.76971875, 'end': 23.90971875, 'confidence': 0.9952150583267212}, {'word': '深', 'start': 23.90971875, 'end': 24.02971875, 'confidence': 0.9548993110656738}, {'word': 'め', 'start': 24.02971875, 'end': 24.22971875, 'confidence': 0.9892219305038452}, {'word': 'ます。', 'start': 24.22971875, 'end': 24.38971875, 'confidence': 0.9906104207038879}], model_info={'diarization_model': 'pyannote/speaker-diarization-3.1', 'transcription_model': 'faster-whisper-small', 'translation_model': 'google_translate'})", "ProcessedSegment(start_time=25.47846875, end_time=25.832843750000002, speaker_id='SPEAKER_00', original_text='見る', original_language='ja', translated_text='See.', confidence_diarization=1.0, confidence_transcription=-0.4798548221588135, confidence_translation=0.8, word_timestamps=[{'word': '見', 'start': 25.47846875, 'end': 25.65846875, 'confidence': 0.5454539060592651}, {'word': 'る', 'start': 25.65846875, 'end': 25.738468750000003, 'confidence': 0.9957653284072876}], model_info={'diarization_model': 'pyannote/speaker-diarization-3.1', 'transcription_model': 'faster-whisper-small', 'translation_model': 'google_translate'})", "ProcessedSegment(start_time=26.204093750000002, end_time=26.65971875, speaker_id='SPEAKER_00', original_text='聞く', original_language='ja', translated_text='Listen.', confidence_diarization=1.0, confidence_transcription=-0.47348871231079104, confidence_translation=0.8, word_timestamps=[{'word': '聞', 'start': 26.204093750000002, 'end': 26.38409375, 'confidence': 0.3832226097583771}, {'word': 'く', 'start': 26.38409375, 'end': 26.524093750000002, 'confidence': 0.9974996447563171}], model_info={'diarization_model': 'pyannote/speaker-diarization-3.1', 'transcription_model': 'faster-whisper-small', 'translation_model': 'google_translate'})", "ProcessedSegment(start_time=26.96346875, end_time=28.617218750000003, speaker_id='SPEAKER_00', original_text='理解するウェブサイトへ', original_language='ja', translated_text='To a website that understands.', confidence_diarization=1.0, confidence_transcription=-0.27092968500577486, confidence_translation=0.8, word_timestamps=[{'word': '理', 'start': 26.96346875, 'end': 27.14346875, 'confidence': 0.4825628995895386}, {'word': '解', 'start': 27.14346875, 'end': 27.36346875, 'confidence': 0.9988553524017334}, {'word': 'する', 'start': 27.36346875, 'end': 27.64346875, 'confidence': 0.9615910649299622}, {'word': 'ウ', 'start': 27.64346875, 'end': 27.903468750000002, 'confidence': 0.4475053548812866}, {'word': 'ェ', 'start': 27.903468750000002, 'end': 28.00346875, 'confidence': 0.9590348601341248}, {'word': 'ブ', 'start': 28.00346875, 'end': 28.08346875, 'confidence': 0.989797830581665}, {'word': 'サ', 'start': 28.08346875, 'end': 28.28346875, 'confidence': 0.9823185205459595}, {'word': 'イ', 'start': 28.28346875, 'end': 28.34346875, 'confidence': 0.998434841632843}, {'word': 'ト', 'start': 28.34346875, 'end': 28.48346875, 'confidence': 0.9974147081375122}, {'word': 'へ', 'start': 28.48346875, 'end': 28.58346875, 'confidence': 0.9876385927200317}], model_info={'diarization_model': 'pyannote/speaker-diarization-3.1', 'transcription_model': 'faster-whisper-small', 'translation_model': 'google_translate'})", "ProcessedSegment(start_time=29.24159375, end_time=31.90784375, speaker_id='SPEAKER_00', original_text='音声メッセージが人の心を動かします', original_language='ja', translated_text=\"And that's what I'm talking about.\", confidence_diarization=1.0, confidence_transcription=-0.23565174551571116, confidence_translation=0.8, word_timestamps=[{'word': '音', 'start': 29.24159375, 'end': 29.42159375, 'confidence': 0.9116391539573669}, {'word': '声', 'start': 29.42159375, 'end': 29.64159375, 'confidence': 0.979734480381012}, {'word': 'メ', 'start': 29.64159375, 'end': 29.78159375, 'confidence': 0.896361768245697}, {'word': 'ッ', 'start': 29.78159375, 'end': 29.86159375, 'confidence': 0.9995806813240051}, {'word': 'セ', 'start': 29.86159375, 'end': 29.96159375, 'confidence': 0.9946938157081604}, {'word': 'ージ', 'start': 29.96159375, 'end': 30.08159375, 'confidence': 0.9994053840637207}, {'word': 'が', 'start': 30.08159375, 'end': 30.28159375, 'confidence': 0.9612740278244019}, {'word': '人', 'start': 30.28159375, 'end': 30.56159375, 'confidence': 0.839630663394928}, {'word': 'の', 'start': 30.56159375, 'end': 30.78159375, 'confidence': 0.9984166622161865}, {'word': '心', 'start': 30.78159375, 'end': 31.00159375, 'confidence': 0.9308077692985535}, {'word': 'を', 'start': 31.00159375, 'end': 31.28159375, 'confidence': 0.9952632188796997}, {'word': '動', 'start': 31.28159375, 'end': 31.42159375, 'confidence': 0.9899610280990601}, {'word': 'か', 'start': 31.42159375, 'end': 31.58159375, 'confidence': 0.9986295700073242}, {'word': 'します', 'start': 31.58159375, 'end': 31.74159375, 'confidence': 0.9892330169677734}], model_info={'diarization_model': 'pyannote/speaker-diarization-3.1', 'transcription_model': 'faster-whisper-small', 'translation_model': 'google_translate'})" ] }