{ "text": "这里是未来预报局AI早报,8月24日。今天关注OpenAI工程师Tibo发布的Codex线流调查结论,以及面向所有付费订阅用户的用量重置预告。OpenAI工程师Tibo在X上发布Codex线流调查更新。他说,团队已定位三类影响用量计算的问题。常绘画多次压缩上下文时,使用图片效率偏低。Computer History功能的高分位段用量偏高,本用于生成绘画标题的功能,消耗超出预期。\nTibo表示,专项小组正在排查,计划次日交付修复。作为修复的一部分,将对所有付费订阅做一次用量全量重置。他在跟帖中表示,用量重置预计于太平洋时间周一下午2点左右进行。他还提到,团队另外发现一种与上述问题无关,可显著提升效率的新方法,将在下周推进。目前修复和用量重置仍是预告,具体额度、规则和覆盖套餐尚未公布。\n以上是今天的AI早报,感谢观看。", "language": "zh", "duration": 60.032, "segments": [ { "id": 0, "seek": 0, "start": 0.0, "end": 29.2, "text": "这里是未来预报局AI早报,8月24日。今天关注OpenAI工程师Tibo发布的Codex线流调查结论,以及面向所有付费订阅用户的用量重置预告。OpenAI工程师Tibo在X上发布Codex线流调查更新。他说,团队已定位三类影响用量计算的问题。常绘画多次压缩上下文时,使用图片效率偏低。Computer History功能的高分位段用量偏高,本用于生成绘画标题的功能,消耗超出预期。", "tokens": [ 50365, 35102, 1541, 29954, 6912, 12501, 226, 49817, 34703, 48698, 21176, 49817, 11, 23, 6939, 7911, 6890, 1543, 12074, 28053, 26432, 45569, 48698, 23323, 29649, 29186, 51, 27776, 28926, 34688, 1546, 34, 1429, 87, 16853, 123, 27854, 8897, 225, 42623, 45641, 7422, 118, 11, 40282, 8833, 24282, 39300, 41469, 18464, 117, 7422, 95, 10034, 227, 9254, 1486, 115, 1546, 9254, 26748, 12624, 34719, 12501, 226, 16846, 1543, 45569, 48698, 23323, 29649, 29186, 51, 27776, 3581, 55, 5708, 28926, 34688, 34, 1429, 87, 16853, 123, 27854, 8897, 225, 42623, 19002, 12560, 1543, 5000, 8090, 11, 3919, 95, 10034, 253, 15701, 12088, 11160, 10960, 22113, 119, 16820, 3858, 235, 9254, 26748, 7422, 94, 19497, 1546, 34069, 1543, 11279, 10115, 246, 27126, 6392, 9487, 5014, 233, 38109, 102, 5708, 4438, 17174, 15729, 11, 22982, 9254, 3919, 122, 16668, 43076, 44866, 7437, 237, 41377, 1543, 14627, 13849, 12486, 32311, 8225, 1546, 12979, 6627, 11160, 28427, 9254, 26748, 7437, 237, 12979, 11, 8802, 9254, 37732, 8244, 11336, 10115, 246, 27126, 162, 3921, 30716, 1546, 32311, 8225, 11, 28837, 4450, 245, 19869, 7781, 5233, 255, 16786, 1543, 51825 ], "temperature": 0, "avg_logprob": -0.05965265502100405, "compression_ratio": 1.2646239554317549, "no_speech_prob": 8.864608699865606e-12, "source": "whisper", "text_original": "这里是未来预报局AI早报,8月24日。今天关注OpenAI工程师Tibo发布的Codex线流调查结论,以及面向所有付费订阅用户的用量重置预告。OpenAI工程师Tibo在X上发布Codex线流调查更新。他说,团队已定位三类影响用量计算的问题。常绘画多次压缩上下文时,使用图片效率偏低。Computer History功能的高分位段用量偏高,本用于生成绘画标题的功能,消耗超出语期。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 1, "seek": 2920, "start": 29.2, "end": 57.04, "text": "Tibo表示,专项小组正在排查,计划次日交付修复。作为修复的一部分,将对所有付费订阅做一次用量全量重置。他在跟帖中表示,用量重置预计于太平洋时间周一下午2点左右进行。他还提到,团队另外发现一种与上述问题无关,可显著提升效率的新方法,将在下周推进。目前修复和用量重置仍是预告,具体额度、规则和覆盖套餐尚未公布。", "tokens": [ 50365, 51, 27776, 40053, 11, 940, 241, 10178, 117, 7322, 10115, 226, 15789, 3581, 44647, 42623, 11, 7422, 94, 2437, 240, 9487, 6890, 28455, 41469, 7792, 106, 1787, 235, 1543, 11914, 13992, 7792, 106, 1787, 235, 1546, 2257, 32174, 11, 45456, 8713, 39300, 41469, 18464, 117, 7422, 95, 10034, 227, 10907, 27505, 9254, 26748, 11319, 26748, 12624, 34719, 1543, 5000, 3581, 9678, 4845, 244, 5975, 40053, 11, 9254, 26748, 12624, 34719, 12501, 226, 7422, 94, 37732, 9455, 16716, 12136, 233, 44848, 37390, 8861, 44237, 17, 12579, 29457, 36700, 8082, 1543, 5000, 14852, 20949, 4511, 11, 3919, 95, 10034, 253, 26202, 28926, 20204, 2257, 39810, 940, 236, 5708, 3316, 108, 34069, 31364, 28053, 11, 4429, 1431, 122, 19382, 20949, 41670, 43076, 44866, 1546, 12560, 9249, 11148, 11, 45456, 3581, 4438, 37390, 33597, 36700, 1543, 39004, 7792, 106, 1787, 235, 12565, 9254, 26748, 12624, 34719, 1550, 235, 1541, 12501, 226, 16846, 11, 39806, 29485, 12501, 251, 13127, 1231, 6758, 226, 2437, 247, 12565, 2888, 228, 5419, 244, 1881, 245, 22217, 238, 1530, 248, 29954, 13545, 34688, 1543, 51757 ], "temperature": 0, "avg_logprob": -0.02398299099354262, "compression_ratio": 1.2861356932153392, "no_speech_prob": 1.6324134752276898e-11, "source": "whisper" }, { "id": 2, "seek": 5704, "start": 57.04, "end": 59.879999999999995, "text": "以上是今天的AI早报,感谢观看。", "tokens": [ 50365, 29497, 1541, 34947, 48698, 21176, 49817, 11, 9709, 11340, 42350, 4200, 1543, 50507 ], "temperature": 0, "avg_logprob": -0.058920033772786456, "compression_ratio": 0.7924528301886793, "no_speech_prob": 2.1927865773152533e-11, "source": "whisper" } ], "transcript_selection": { "schema": "videonote.transcript_selection.v1", "status": "SELECTED", "requested_language": "zh", "duration": 60.032, "reference_path": null, "reference_kind": null, "reference_language": null, "selected_source": "whisper_segmented", "decision_reason": "source_unavailable_or_untrusted_whisper_passed", "whisper_rejected": false, "source_profile": { "segment_count": 0, "duration_coverage": 0.0, "usable_full": false, "usable_partial": false, "hallucination_region_count": 0, "hallucination_regions": [], "language": { "requested": "zh", "matched": false } }, "whisper_profile": { "segment_count": 3, "multilingual": false, "first_start": 0.0, "final_end": 59.88, "duration_coverage": 0.997468, "maximum_gap_seconds": 0.0, "inverted_count": 0, "language": { "requested": "zh", "matched": true, "latin": 54, "han": 273, "kana": 0, "hangul": 0 }, "hallucination_region_count": 0, "hallucination_regions": [], "usable_full": true, "usable_partial": true }, "comparison": { "window_seconds": 30.0, "compared_window_count": 0, "median_score": 0.0, "divergent_window_count": 0, "divergent_ratio": 0.0, "divergent_windows": [] }, "timestamp_alignment": { "applied": false, "estimated_offset_seconds": 0.0, "matched_segment_count": 0, "match_scores": [] }, "caption_timing_repair": null, "filled_source_profile": null, "whisper_duration_guard": { "media_duration": 60.032, "rejected_segment_count": 0, "rejected_segments": [] } }, "presegmentation_correction": { "applied": true, "correction_count": 1, "deterministic_fix_count": 0 } }