{
  "text": "大家好,我是Cade。Deepseek V4 Flash 正式版API今天开始供测了。官方介绍agent的能力大幅增强。基准测试远超V4 Pro Preview。大家还记得V4 Flash是一个比较小的模型吗?正式版V4 Flash还原生支持Responses API格式,并且针对性的适配Codex。正式版的Flash它的尺寸和Preview是一致的,仅重新进行了后训练。这是官方介绍的揭露Codex的教程。\nMac电脑就输入这行命令。如果Windows是这行命令。Mac电脑上运行了这行命令之后,它就会出现三个选择。让我选,我选择第一个修改Codex的配置。接着就需要填写Deepseek的API Key。接着就会自动备份写入更新。它改写了我模型的默认模型,还有将努力程度改成了Hi。我原来的Config文件里面呢,有些内容它也\n给我覆盖掉了。那这部分是写入后的内容。再接着自动校验,显示安装完成。如果你想要恢复的话,可以再次运行一开始的命令。选择三呢,就会恢复到之前的配置。目前按照官方的教程,配置好Flash之后,GPT的客户端呢,它就会统一显示自定义。\n不同的推理程度的。如果说你选择的是Low和Medium,那就自动显示到Deepseek里的Hi推理模式。如果你选择是GPT的XHi,那它也会显示到Max的推理程度。官方说,正式版的V4Flash\n比之前越南版的Pro还要强。那我们可以再回顾一下V4Flash的参数。它是284B,激活13B。目前呢,它还是不支持图像,音频,视频输入。我们会看到TOMEL BENCH 2.1\nV4Flash已经和质朴的5.2是得分一致了。而在Deepseek的基准上,它的得分居然是要超过5.2的。而在人类最后的测试方面呢,V4Flash得分不如质朴5.2。我测试了V4Flash,我发现它挺有意思的。它虽然是不能视图,但是它在运行中,它会经常使用像素统计和渲染进行诊断。\n也就是用数字代替眼睛去检查画面好不好。现在就看它的实测,这是它做的清明上合图。右侧呢,有菜单。我先点击一下虹桥全景。然后再看一下桥上的人群。这些呢,都是它程序化生成的。没有引入外部的3D资产。再看一下桥下的船工。然后,码头装卸这里被挡住了。\n9楼2层也是被挡住了。\n再来看一下河岸长卷。\n城门俯瞰。\n我们可以看到它很多房屋,屋顶都是没有构建好的。\n做得非常非常简约。\n对比我最近测试过的无论是K3,或者是GPT模型。\n那,Winz Flash在这个任务上完成的是非常一般的。\n但是又考虑到它的规模的话,那我觉得它其实已经完成得非常好了。\n那,点击一下沿河漫游,可以看到是比较卡顿的。\n再看它生成的桌面行星仪。\n它生成的各个行星,材料的质感还行。\n但是画面里,有非常多的错误。\n而且这些星球也不转了。\n再看它生成的苏州博物馆。\n对于这样一个模型,我认为它做出来的效果,尤其是用石片做的一些假衫。\n它做的还行,比我想象中的好。\n再看它生成的上海地多边形城市地标沙河。\n展示了不同的建筑。\n人民广场。\n博物馆。\n金安寺陆家嘴。\n世博会博物馆。\n对上海地里比较熟悉的朋友,可以在弹幕里回复一下,看看是不是生成的比较好。\n河道的话也做的太笔直了。\n再看一下它做的黑白线稿互动咖啡馆。\n那我是之前在B站上有看到用户用K3做一个线稿的,感觉不错。\n后来我就参考它的idea。\n那现在让Deepseek V4 Flash来做这样的一个互动。\n点击画面里的不同元素是可以有互动的。\n比如说刚刚这里窗户外有下雨。\n然后这里点击咖啡豆呢,咖啡豆有溢出。\n咖啡机点击之后,它会有蒸腾沸腾的效果。\n那点击人物,它这里会有插杯子的动作。\n点击咖啡杯呢,它也会做一个拉花的效果。\n菜单也可以交互。\n然后这里也有一些植物。\n包括人物,这里的椅子也是可以移动。\n桌子也是可以摇动。\n整体我感觉还不错。\n再给大家看一下我用OPS5生成的效果。\n。\n。\n。\nOPS5呢,它是我提示了它两次。\n它一开始呢,对画面里做的交互不是很丰富。\n所以后来我又提示让它去优化一下。\n而且它的整体速度是非常慢的。\n当然它花费的价格也是远远高于V4 Flash的。\n其实我还有两个任务呢,让V4 Flash在运行着。\n那它们已经运行了一个多小时了,还没有结束。\n其实我在日常里呢,还是会经常用到V4 Flash的。\n比如说现在看到的这个应用呢,\n我就是在半小时之内,让Glock 4.5生成的。\n揭露的是Dipsic V4 Flash和Pro模型。\n因为它本身是EM的上下文窗口。\n而且它的生成速度很快,价格很便宜。\n所以我会把它揭露在我自己用的一些App里,\n或者是用于翻译的插件里。\n都是速度、质量都是非常好的。\nV4 Flash的应用场景还是非常多的。\n但就编程而言,因为我们现在也是有大量的,\n更大规模的前沿模型可以使用。\n无论是K3,Office,或者是GPT-Sore模型,\n或者是Glock 4.5。\n那我个人认为的话,\nV4 Flash正式版在编码上还是弱了一些。\n简单的,我让它执行OK,\n但是更复杂的。\n或者说你更想OneShot来去生成一个比较好的结果的话,\n可能我会更倾向于用GPT模型,\n或者是Glock模型。\n以上就是今天的所有分享,\n希望我的视频对你有帮助,\n我们下次再见。",
  "language": "zh",
  "duration": 503.965938,
  "segments": [
    {
      "id": 0,
      "seek": 0,
      "start": 0.0,
      "end": 29.3,
      "text": "大家好,我是Cade。Deepseek V4 Flash 正式版API今天开始供测了。官方介绍agent的能力大幅增强。基准测试远超V4 Pro Preview。大家还记得V4 Flash是一个比较小的模型吗?正式版V4 Flash还原生支持Responses API格式,并且针对性的适配Codex。正式版的Flash它的尺寸和Preview是一致的,仅重新进行了后训练。这是官方介绍的揭露Codex的教程。",
      "tokens": [
        50365,
        15248,
        11,
        15914,
        34,
        762,
        1543,
        11089,
        595,
        405,
        916,
        691,
        19,
        20232,
        220,
        15789,
        27584,
        42096,
        4715,
        40,
        12074,
        45213,
        3254,
        249,
        11038,
        233,
        2289,
        1543,
        31929,
        9249,
        30312,
        10115,
        235,
        559,
        317,
        1546,
        8225,
        13486,
        3582,
        3509,
        227,
        24228,
        252,
        5702,
        118,
        1543,
        26008,
        6336,
        228,
        11038,
        233,
        5233,
        243,
        3316,
        250,
        19869,
        53,
        19,
        1705,
        6001,
        1759,
        1543,
        6868,
        14852,
        34756,
        5916,
        53,
        19,
        20232,
        1541,
        20182,
        11706,
        9830,
        225,
        7322,
        1546,
        41908,
        39823,
        14769,
        30,
        15789,
        27584,
        42096,
        53,
        19,
        20232,
        14852,
        19683,
        8244,
        35488,
        49,
        13361,
        892,
        279,
        9362,
        30921,
        27584,
        11,
        3509,
        114,
        20334,
        25153,
        230,
        8713,
        21686,
        1546,
        2215,
        224,
        38846,
        34,
        1429,
        87,
        1543,
        15789,
        27584,
        42096,
        1546,
        25680,
        1299,
        45224,
        1530,
        118,
        4510,
        116,
        12565,
        42844,
        1759,
        1541,
        2257,
        6784,
        112,
        1546,
        11,
        1550,
        227,
        12624,
        12560,
        36700,
        46606,
        13547,
        7422,
        255,
        10115,
        225,
        1543,
        27455,
        31929,
        9249,
        30312,
        10115,
        235,
        1546,
        11673,
        255,
        18594,
        110,
        34,
        1429,
        87,
        1546,
        21936,
        29649,
        1543,
        51830
      ],
      "temperature": 0,
      "avg_logprob": -0.1968006225953619,
      "compression_ratio": 1.2373887240356083,
      "no_speech_prob": 9.51205520599574e-12,
      "source": "whisper"
    },
    {
      "id": 1,
      "seek": 2930,
      "start": 29.3,
      "end": 59.28,
      "text": "Mac电脑就输入这行命令。如果Windows是这行命令。Mac电脑上运行了这行命令之后,它就会出现三个选择。让我选,我选择第一个修改Codex的配置。接着就需要填写Deepseek的API Key。接着就会自动备份写入更新。它改写了我模型的默认模型,还有将努力程度改成了Hi。我原来的Config文件里面呢,有些内容它也",
      "tokens": [
        50365,
        34355,
        42182,
        27067,
        239,
        3111,
        9830,
        241,
        14028,
        5562,
        8082,
        25236,
        49061,
        1543,
        13119,
        45813,
        1509,
        1541,
        5562,
        8082,
        25236,
        49061,
        1543,
        34355,
        42182,
        27067,
        239,
        5708,
        3316,
        238,
        46606,
        5562,
        8082,
        25236,
        49061,
        9574,
        13547,
        11,
        11284,
        3111,
        12949,
        7781,
        20204,
        10960,
        7549,
        2215,
        231,
        6852,
        102,
        1543,
        33650,
        1654,
        2215,
        231,
        11,
        1654,
        2215,
        231,
        6852,
        102,
        18049,
        7549,
        7792,
        106,
        34490,
        34,
        1429,
        87,
        1546,
        38846,
        34719,
        1543,
        14468,
        20708,
        3111,
        35748,
        13331,
        104,
        5676,
        247,
        11089,
        595,
        405,
        916,
        1546,
        4715,
        40,
        12759,
        1543,
        14468,
        20708,
        3111,
        12949,
        9722,
        34961,
        1787,
        229,
        36266,
        5676,
        247,
        14028,
        19002,
        12560,
        1543,
        11284,
        34490,
        5676,
        247,
        2289,
        1654,
        41908,
        39823,
        1546,
        6173,
        246,
        7422,
        97,
        41908,
        39823,
        11,
        35091,
        45456,
        5087,
        103,
        13486,
        29649,
        13127,
        34490,
        11336,
        2289,
        17155,
        1543,
        1654,
        19683,
        6912,
        1546,
        9838,
        20646,
        17174,
        20485,
        15759,
        8833,
        6240,
        11,
        2412,
        13824,
        34742,
        25750,
        11284,
        6404,
        51864
      ],
      "temperature": 0,
      "avg_logprob": -0.10001654406778174,
      "compression_ratio": 1.2420382165605095,
      "no_speech_prob": 1.5276724329993385e-11,
      "source": "whisper"
    },
    {
      "id": 2,
      "seek": 5928,
      "start": 59.28,
      "end": 89.28,
      "text": "给我覆盖掉了。那这部分是写入后的内容。再接着自动校验,显示安装完成。如果你想要恢复的话,可以再次运行一开始的命令。选择三呢,就会恢复到之前的配置。目前按照官方的教程,配置好Flash之后,GPT的客户端呢,它就会统一显示自定义。",
      "tokens": [
        50365,
        49076,
        2888,
        228,
        5419,
        244,
        29327,
        2289,
        1543,
        4184,
        5562,
        32174,
        1541,
        5676,
        247,
        14028,
        13547,
        1546,
        34742,
        25750,
        1543,
        8623,
        14468,
        20708,
        9722,
        34961,
        49684,
        49657,
        234,
        11,
        1431,
        122,
        25696,
        16206,
        8083,
        227,
        41509,
        1543,
        13119,
        38386,
        4275,
        11845,
        95,
        1787,
        235,
        44575,
        11,
        6723,
        8623,
        9487,
        3316,
        238,
        8082,
        2257,
        45213,
        1546,
        25236,
        49061,
        1543,
        2215,
        231,
        6852,
        102,
        10960,
        6240,
        11,
        3111,
        12949,
        11845,
        95,
        1787,
        235,
        4511,
        32442,
        1546,
        38846,
        34719,
        1543,
        39004,
        26613,
        32150,
        31929,
        9249,
        1546,
        21936,
        29649,
        11,
        38846,
        34719,
        2131,
        25680,
        1299,
        9574,
        13547,
        11,
        38,
        47,
        51,
        1546,
        32316,
        1486,
        115,
        11957,
        107,
        6240,
        11,
        11284,
        3111,
        12949,
        10115,
        253,
        2257,
        1431,
        122,
        25696,
        9722,
        12088,
        2930,
        231,
        1543
      ],
      "temperature": 0,
      "avg_logprob": -0.09424163881412223,
      "compression_ratio": 1.2076923076923076,
      "no_speech_prob": 1.5762920482775833e-11,
      "source": "whisper"
    },
    {
      "id": 3,
      "seek": 8928,
      "start": 89.28,
      "end": 105.28,
      "text": "不同的推理程度的。如果说你选择的是Low和Medium,那就自动显示到Deepseek里的Hi推理模式。如果你选择是GPT的XHi,那它也会显示到Max的推理程度。官方说,正式版的V4Flash",
      "tokens": [
        50365,
        47123,
        1546,
        33597,
        13876,
        29649,
        13127,
        1546,
        1543,
        13119,
        8090,
        2166,
        2215,
        231,
        6852,
        102,
        24620,
        43,
        305,
        12565,
        42954,
        2197,
        11,
        46887,
        9722,
        34961,
        1431,
        122,
        25696,
        4511,
        11089,
        595,
        405,
        916,
        15759,
        1546,
        17155,
        33597,
        13876,
        41908,
        27584,
        1543,
        45669,
        2215,
        231,
        6852,
        102,
        1541,
        38,
        47,
        51,
        1546,
        55,
        17155,
        11,
        4184,
        11284,
        6404,
        12949,
        1431,
        122,
        25696,
        4511,
        36025,
        1546,
        33597,
        13876,
        29649,
        13127,
        1543,
        31929,
        9249,
        8090,
        11,
        15789,
        27584,
        42096,
        1546,
        53,
        19,
        25680,
        1299,
        51165
      ],
      "temperature": 0,
      "avg_logprob": -0.14253543404971852,
      "compression_ratio": 1.1944444444444444,
      "no_speech_prob": 1.8698712717091226e-11,
      "source": "whisper"
    },
    {
      "id": 4,
      "seek": 10528,
      "start": 105.28,
      "end": 122.28,
      "text": "比之前越南版的Pro还要强。那我们可以再回顾一下V4Flash的参数。它是284B,激活13B。目前呢,它还是不支持图像,音频,视频输入。我们会看到TOMEL BENCH 2.1",
      "tokens": [
        50365,
        11706,
        32442,
        25761,
        31021,
        42096,
        1546,
        12681,
        14852,
        4275,
        5702,
        118,
        1543,
        4184,
        15003,
        6723,
        8623,
        8350,
        10178,
        122,
        8861,
        53,
        19,
        25680,
        1299,
        1546,
        2129,
        224,
        33188,
        1543,
        11284,
        1541,
        11205,
        19,
        33,
        11,
        26373,
        222,
        25956,
        7668,
        33,
        1543,
        39004,
        6240,
        11,
        11284,
        45726,
        1960,
        35488,
        3919,
        122,
        12760,
        11,
        18034,
        39752,
        11,
        40656,
        39752,
        9830,
        241,
        14028,
        1543,
        15003,
        12949,
        18032,
        51,
        5251,
        3158,
        31613,
        5462,
        568,
        13,
        16,
        51215
      ],
      "temperature": 0,
      "avg_logprob": -0.17442196293881065,
      "compression_ratio": 0.975,
      "no_speech_prob": 1.8066225598856178e-11,
      "source": "whisper"
    },
    {
      "id": 5,
      "seek": 12228,
      "start": 122.28,
      "end": 150.28,
      "text": "V4Flash已经和质朴的5.2是得分一致了。而在Deepseek的基准上,它的得分居然是要超过5.2的。而在人类最后的测试方面呢,V4Flash得分不如质朴5.2。我测试了V4Flash,我发现它挺有意思的。它虽然是不能视图,但是它在运行中,它会经常使用像素统计和渲染进行诊断。",
      "tokens": [
        50365,
        53,
        19,
        25680,
        1299,
        49161,
        12565,
        18464,
        101,
        1474,
        112,
        1546,
        20,
        13,
        17,
        1541,
        5916,
        6627,
        2257,
        6784,
        112,
        2289,
        1543,
        11070,
        3581,
        11089,
        595,
        405,
        916,
        1546,
        26008,
        6336,
        228,
        5708,
        11,
        45224,
        5916,
        6627,
        9636,
        227,
        5823,
        1541,
        4275,
        19869,
        16866,
        20,
        13,
        17,
        1546,
        1543,
        11070,
        3581,
        4035,
        22113,
        119,
        8661,
        13547,
        1546,
        11038,
        233,
        5233,
        243,
        39604,
        6240,
        11,
        53,
        19,
        25680,
        1299,
        5916,
        6627,
        1960,
        8238,
        18464,
        101,
        1474,
        112,
        20,
        13,
        17,
        1543,
        1654,
        11038,
        233,
        5233,
        243,
        2289,
        53,
        19,
        25680,
        1299,
        11,
        1654,
        28926,
        20204,
        11284,
        41046,
        2412,
        16697,
        1546,
        1543,
        11284,
        12026,
        121,
        5823,
        1541,
        28590,
        40656,
        3919,
        122,
        11,
        11189,
        11284,
        3581,
        3316,
        238,
        8082,
        5975,
        11,
        11284,
        12949,
        30276,
        11279,
        22982,
        9254,
        12760,
        37204,
        10115,
        253,
        7422,
        94,
        12565,
        9592,
        110,
        46568,
        36700,
        8082,
        5233,
        232,
        4307,
        255,
        1543,
        51765
      ],
      "temperature": 0,
      "avg_logprob": -0.0925372288144868,
      "compression_ratio": 1.2324723247232472,
      "no_speech_prob": 1.5218510479586556e-11,
      "source": "whisper"
    },
    {
      "id": 6,
      "seek": 15028,
      "start": 150.28,
      "end": 178.46,
      "text": "也就是用数字代替眼睛去检查画面好不好。现在就看它的实测,这是它做的清明上合图。右侧呢,有菜单。我先点击一下虹桥全景。然后再看一下桥上的人群。这些呢,都是它程序化生成的。没有引入外部的3D资产。再看一下桥下的船工。然后,码头装卸这里被挡住了。",
      "tokens": [
        50365,
        6404,
        5620,
        9254,
        33188,
        22381,
        19105,
        9531,
        123,
        25281,
        14221,
        249,
        6734,
        16407,
        222,
        42623,
        27126,
        8833,
        34291,
        1543,
        25040,
        3111,
        4200,
        45224,
        24726,
        11038,
        233,
        11,
        27455,
        11284,
        10907,
        1546,
        21784,
        11100,
        5708,
        14245,
        3919,
        122,
        1543,
        22060,
        3254,
        100,
        6240,
        11,
        2412,
        33025,
        47446,
        1543,
        1654,
        10108,
        12579,
        6336,
        119,
        8861,
        12026,
        117,
        21416,
        98,
        11319,
        50218,
        1543,
        26636,
        8623,
        28324,
        21416,
        98,
        5708,
        29979,
        7248,
        97,
        1543,
        5562,
        13824,
        6240,
        11,
        22796,
        11284,
        29649,
        6346,
        237,
        23756,
        8244,
        11336,
        1546,
        1543,
        17944,
        41301,
        14028,
        12022,
        13470,
        1546,
        18,
        35,
        5266,
        226,
        1369,
        100,
        1543,
        8623,
        28324,
        21416,
        98,
        4438,
        1546,
        10256,
        117,
        23323,
        1543,
        26636,
        11,
        23230,
        223,
        39862,
        8083,
        227,
        5322,
        116,
        35102,
        23238,
        8501,
        94,
        21632,
        2289,
        1543,
        51774
      ],
      "temperature": 0,
      "avg_logprob": -0.10009209940752645,
      "compression_ratio": 1.2936802973977695,
      "no_speech_prob": 1.7714305716731715e-11,
      "source": "whisper"
    },
    {
      "id": 7,
      "seek": 17846,
      "start": 178.46,
      "end": 182.54000000000002,
      "text": "9楼2层也是被挡住了。",
      "tokens": [
        50365,
        24,
        162,
        2429,
        17,
        9636,
        224,
        22021,
        23238,
        8501,
        94,
        21632,
        2289,
        1543,
        50569
      ],
      "temperature": 0,
      "avg_logprob": -0.23623701644270387,
      "compression_ratio": 1.0632911392405062,
      "no_speech_prob": 1.6049831602638065e-11,
      "source": "whisper"
    },
    {
      "id": 8,
      "seek": 17846,
      "start": 183.70000000000002,
      "end": 185.72,
      "text": "再来看一下河岸长卷。",
      "tokens": [
        50627,
        8623,
        6912,
        28324,
        3308,
        111,
        23182,
        116,
        32271,
        5322,
        115,
        1543,
        50728
      ],
      "temperature": 0,
      "avg_logprob": -0.23623701644270387,
      "compression_ratio": 1.0632911392405062,
      "no_speech_prob": 1.6049831602638065e-11,
      "source": "whisper"
    },
    {
      "id": 9,
      "seek": 17846,
      "start": 186.8,
      "end": 188.10000000000002,
      "text": "城门俯瞰。",
      "tokens": [
        50782,
        45277,
        8259,
        101,
        7792,
        107,
        36733,
        108,
        1543,
        50847
      ],
      "temperature": 0,
      "avg_logprob": -0.23623701644270387,
      "compression_ratio": 1.0632911392405062,
      "no_speech_prob": 1.6049831602638065e-11,
      "source": "whisper"
    },
    {
      "id": 10,
      "seek": 17846,
      "start": 189.16,
      "end": 193.74,
      "text": "我们可以看到它很多房屋,屋顶都是没有构建好的。",
      "tokens": [
        50900,
        15003,
        6723,
        18032,
        11284,
        20778,
        38242,
        9636,
        233,
        11,
        9636,
        233,
        10178,
        114,
        22796,
        17944,
        7360,
        226,
        34157,
        20715,
        1543,
        51129
      ],
      "temperature": 0,
      "avg_logprob": -0.23623701644270387,
      "compression_ratio": 1.0632911392405062,
      "no_speech_prob": 1.6049831602638065e-11,
      "source": "whisper"
    },
    {
      "id": 11,
      "seek": 17846,
      "start": 194.36,
      "end": 195.92000000000002,
      "text": "做得非常非常简约。",
      "tokens": [
        51160,
        10907,
        5916,
        14392,
        14392,
        11249,
        222,
        16853,
        99,
        1543,
        51238
      ],
      "temperature": 0,
      "avg_logprob": -0.23623701644270387,
      "compression_ratio": 1.0632911392405062,
      "no_speech_prob": 1.6049831602638065e-11,
      "source": "whisper"
    },
    {
      "id": 12,
      "seek": 19592,
      "start": 195.92,
      "end": 202.92,
      "text": "对比我最近测试过的无论是K3,或者是GPT模型。",
      "tokens": [
        50365,
        8713,
        11706,
        1654,
        46777,
        11038,
        233,
        5233,
        243,
        16866,
        1546,
        31364,
        7422,
        118,
        1541,
        42,
        18,
        11,
        31148,
        1541,
        38,
        47,
        51,
        41908,
        39823,
        1543,
        50715
      ],
      "temperature": 0,
      "avg_logprob": -0.19363283466648412,
      "compression_ratio": 1.0714285714285714,
      "no_speech_prob": 1.6268120531237606e-11,
      "source": "whisper"
    },
    {
      "id": 13,
      "seek": 19592,
      "start": 203.66,
      "end": 207.44,
      "text": "那,Winz Flash在这个任务上完成的是非常一般的。",
      "tokens": [
        50752,
        4184,
        11,
        54,
        259,
        89,
        20232,
        3581,
        15368,
        26443,
        5087,
        94,
        5708,
        41509,
        24620,
        14392,
        2257,
        49640,
        1546,
        1543,
        50941
      ],
      "temperature": 0,
      "avg_logprob": -0.19363283466648412,
      "compression_ratio": 1.0714285714285714,
      "no_speech_prob": 1.6268120531237606e-11,
      "source": "whisper"
    },
    {
      "id": 14,
      "seek": 19592,
      "start": 207.76,
      "end": 212.92,
      "text": "但是又考虑到它的规模的话,那我觉得它其实已经完成得非常好了。",
      "tokens": [
        50957,
        11189,
        17047,
        26504,
        12026,
        239,
        4511,
        45224,
        6758,
        226,
        41908,
        44575,
        11,
        4184,
        45681,
        11284,
        41646,
        49161,
        41509,
        5916,
        14392,
        12621,
        1543,
        51215
      ],
      "temperature": 0,
      "avg_logprob": -0.19363283466648412,
      "compression_ratio": 1.0714285714285714,
      "no_speech_prob": 1.6268120531237606e-11,
      "source": "whisper"
    },
    {
      "id": 15,
      "seek": 21292,
      "start": 212.92,
      "end": 218.17999999999998,
      "text": "那,点击一下沿河漫游,可以看到是比较卡顿的。",
      "tokens": [
        50365,
        4184,
        11,
        12579,
        6336,
        119,
        8861,
        3308,
        123,
        3308,
        111,
        14065,
        104,
        9592,
        116,
        11,
        6723,
        18032,
        1541,
        11706,
        9830,
        225,
        32681,
        10178,
        123,
        1546,
        1543,
        50628
      ],
      "temperature": 0,
      "avg_logprob": -0.14555883407592773,
      "compression_ratio": 1.2673267326732673,
      "no_speech_prob": 1.3689181732612354e-11,
      "source": "whisper"
    },
    {
      "id": 16,
      "seek": 21292,
      "start": 218.85999999999999,
      "end": 220.38,
      "text": "再看它生成的桌面行星仪。",
      "tokens": [
        50662,
        8623,
        4200,
        11284,
        8244,
        11336,
        1546,
        162,
        5475,
        8833,
        8082,
        20682,
        1550,
        103,
        1543,
        50738
      ],
      "temperature": 0,
      "avg_logprob": -0.14555883407592773,
      "compression_ratio": 1.2673267326732673,
      "no_speech_prob": 1.3689181732612354e-11,
      "source": "whisper"
    },
    {
      "id": 17,
      "seek": 21292,
      "start": 220.88,
      "end": 223.92,
      "text": "它生成的各个行星,材料的质感还行。",
      "tokens": [
        50763,
        11284,
        8244,
        11336,
        1546,
        17516,
        7549,
        8082,
        20682,
        11,
        4422,
        238,
        33404,
        1546,
        18464,
        101,
        9709,
        14852,
        8082,
        1543,
        50915
      ],
      "temperature": 0,
      "avg_logprob": -0.14555883407592773,
      "compression_ratio": 1.2673267326732673,
      "no_speech_prob": 1.3689181732612354e-11,
      "source": "whisper"
    },
    {
      "id": 18,
      "seek": 21292,
      "start": 224.44,
      "end": 227.26,
      "text": "但是画面里,有非常多的错误。",
      "tokens": [
        50941,
        11189,
        27126,
        8833,
        15759,
        11,
        2412,
        14392,
        6392,
        1546,
        29900,
        5233,
        107,
        1543,
        51082
      ],
      "temperature": 0,
      "avg_logprob": -0.14555883407592773,
      "compression_ratio": 1.2673267326732673,
      "no_speech_prob": 1.3689181732612354e-11,
      "source": "whisper"
    },
    {
      "id": 19,
      "seek": 21292,
      "start": 227.44,
      "end": 229.06,
      "text": "而且这些星球也不转了。",
      "tokens": [
        51091,
        22942,
        5562,
        13824,
        20682,
        28533,
        6404,
        1960,
        17819,
        105,
        2289,
        1543,
        51172
      ],
      "temperature": 0,
      "avg_logprob": -0.14555883407592773,
      "compression_ratio": 1.2673267326732673,
      "no_speech_prob": 1.3689181732612354e-11,
      "source": "whisper"
    },
    {
      "id": 20,
      "seek": 21292,
      "start": 229.44,
      "end": 231.44,
      "text": "再看它生成的苏州博物馆。",
      "tokens": [
        51191,
        8623,
        4200,
        11284,
        8244,
        11336,
        1546,
        13736,
        237,
        46974,
        5322,
        248,
        23516,
        11748,
        228,
        1543,
        51291
      ],
      "temperature": 0,
      "avg_logprob": -0.14555883407592773,
      "compression_ratio": 1.2673267326732673,
      "no_speech_prob": 1.3689181732612354e-11,
      "source": "whisper"
    },
    {
      "id": 21,
      "seek": 23144,
      "start": 231.44,
      "end": 239.42,
      "text": "对于这样一个模型,我认为它做出来的效果,尤其是用石片做的一些假衫。",
      "tokens": [
        50365,
        8713,
        37732,
        21209,
        20182,
        41908,
        39823,
        11,
        1654,
        7422,
        97,
        13992,
        11284,
        10907,
        44561,
        1546,
        43076,
        9319,
        11,
        1530,
        97,
        9572,
        1541,
        9254,
        36783,
        16668,
        10907,
        1546,
        38515,
        31706,
        9890,
        104,
        1543,
        50764
      ],
      "temperature": 0,
      "avg_logprob": -0.21216484478541783,
      "compression_ratio": 1.2081447963800904,
      "no_speech_prob": 1.680931956127285e-11,
      "source": "whisper"
    },
    {
      "id": 22,
      "seek": 23144,
      "start": 240.35999999999999,
      "end": 244.07999999999998,
      "text": "它做的还行,比我想象中的好。",
      "tokens": [
        50811,
        11284,
        10907,
        1546,
        14852,
        8082,
        11,
        11706,
        25246,
        45007,
        5975,
        1546,
        2131,
        1543,
        50997
      ],
      "temperature": 0,
      "avg_logprob": -0.21216484478541783,
      "compression_ratio": 1.2081447963800904,
      "no_speech_prob": 1.680931956127285e-11,
      "source": "whisper"
    },
    {
      "id": 23,
      "seek": 23144,
      "start": 244.96,
      "end": 248.26,
      "text": "再看它生成的上海地多边形城市地标沙河。",
      "tokens": [
        51041,
        8623,
        4200,
        11284,
        8244,
        11336,
        1546,
        5708,
        20078,
        10928,
        6392,
        40503,
        30900,
        45277,
        27261,
        10928,
        162,
        3921,
        3308,
        247,
        3308,
        111,
        1543,
        51206
      ],
      "temperature": 0,
      "avg_logprob": -0.21216484478541783,
      "compression_ratio": 1.2081447963800904,
      "no_speech_prob": 1.680931956127285e-11,
      "source": "whisper"
    },
    {
      "id": 24,
      "seek": 23144,
      "start": 250.54,
      "end": 252.28,
      "text": "展示了不同的建筑。",
      "tokens": [
        51320,
        43491,
        25696,
        2289,
        47123,
        1546,
        34157,
        7973,
        239,
        1543,
        51407
      ],
      "temperature": 0,
      "avg_logprob": -0.21216484478541783,
      "compression_ratio": 1.2081447963800904,
      "no_speech_prob": 1.680931956127285e-11,
      "source": "whisper"
    },
    {
      "id": 25,
      "seek": 23144,
      "start": 257.06,
      "end": 258.15999999999997,
      "text": "人民广场。",
      "tokens": [
        51646,
        44894,
        3509,
        123,
        50255,
        1543,
        51701
      ],
      "temperature": 0,
      "avg_logprob": -0.21216484478541783,
      "compression_ratio": 1.2081447963800904,
      "no_speech_prob": 1.680931956127285e-11,
      "source": "whisper"
    },
    {
      "id": 26,
      "seek": 23144,
      "start": 258.82,
      "end": 259.36,
      "text": "博物馆。",
      "tokens": [
        51734,
        5322,
        248,
        23516,
        11748,
        228,
        1543,
        51761
      ],
      "temperature": 0,
      "avg_logprob": -0.21216484478541783,
      "compression_ratio": 1.2081447963800904,
      "no_speech_prob": 1.680931956127285e-11,
      "source": "whisper"
    },
    {
      "id": 27,
      "seek": 23144,
      "start": 260.0,
      "end": 261.22,
      "text": "金安寺陆家嘴。",
      "tokens": [
        51793,
        19117,
        16206,
        4510,
        118,
        8842,
        228,
        5155,
        10756,
        112,
        1543,
        51854
      ],
      "temperature": 0,
      "avg_logprob": -0.21216484478541783,
      "compression_ratio": 1.2081447963800904,
      "no_speech_prob": 1.680931956127285e-11,
      "source": "whisper"
    },
    {
      "id": 28,
      "seek": 26144,
      "start": 261.44,
      "end": 262.84,
      "text": "世博会博物馆。",
      "tokens": [
        50365,
        17484,
        5322,
        248,
        12949,
        5322,
        248,
        23516,
        11748,
        228,
        1543,
        50435
      ],
      "temperature": 0,
      "avg_logprob": -0.17641683058305221,
      "compression_ratio": 1.2290076335877862,
      "no_speech_prob": 1.254823155105722e-11,
      "source": "whisper"
    },
    {
      "id": 29,
      "seek": 26144,
      "start": 265.04,
      "end": 273.42,
      "text": "对上海地里比较熟悉的朋友,可以在弹幕里回复一下,看看是不是生成的比较好。",
      "tokens": [
        50545,
        8713,
        5708,
        20078,
        10928,
        15759,
        11706,
        9830,
        225,
        21805,
        253,
        14696,
        231,
        1546,
        19828,
        11,
        6723,
        3581,
        5702,
        117,
        40211,
        15759,
        8350,
        1787,
        235,
        8861,
        11,
        22884,
        23034,
        8244,
        11336,
        1546,
        11706,
        9830,
        225,
        2131,
        1543,
        50964
      ],
      "temperature": 0,
      "avg_logprob": -0.17641683058305221,
      "compression_ratio": 1.2290076335877862,
      "no_speech_prob": 1.254823155105722e-11,
      "source": "whisper"
    },
    {
      "id": 30,
      "seek": 26144,
      "start": 273.68,
      "end": 276.12,
      "text": "河道的话也做的太笔直了。",
      "tokens": [
        50977,
        3308,
        111,
        6025,
        44575,
        6404,
        10907,
        1546,
        9455,
        5437,
        242,
        16186,
        2289,
        1543,
        51099
      ],
      "temperature": 0,
      "avg_logprob": -0.17641683058305221,
      "compression_ratio": 1.2290076335877862,
      "no_speech_prob": 1.254823155105722e-11,
      "source": "whisper"
    },
    {
      "id": 31,
      "seek": 26144,
      "start": 276.96,
      "end": 280.28,
      "text": "再看一下它做的黑白线稿互动咖啡馆。",
      "tokens": [
        51141,
        8623,
        28324,
        11284,
        10907,
        1546,
        32141,
        13558,
        16853,
        123,
        10415,
        123,
        1369,
        240,
        34961,
        8975,
        244,
        3284,
        94,
        11748,
        228,
        1543,
        51307
      ],
      "temperature": 0,
      "avg_logprob": -0.17641683058305221,
      "compression_ratio": 1.2290076335877862,
      "no_speech_prob": 1.254823155105722e-11,
      "source": "whisper"
    },
    {
      "id": 32,
      "seek": 26144,
      "start": 280.64,
      "end": 288.28,
      "text": "那我是之前在B站上有看到用户用K3做一个线稿的,感觉不错。",
      "tokens": [
        51325,
        4184,
        15914,
        32442,
        3581,
        33,
        34155,
        5708,
        2412,
        18032,
        9254,
        1486,
        115,
        9254,
        42,
        18,
        10907,
        20182,
        16853,
        123,
        10415,
        123,
        1546,
        11,
        9709,
        24447,
        1960,
        29900,
        1543,
        51707
      ],
      "temperature": 0,
      "avg_logprob": -0.17641683058305221,
      "compression_ratio": 1.2290076335877862,
      "no_speech_prob": 1.254823155105722e-11,
      "source": "whisper"
    },
    {
      "id": 33,
      "seek": 26144,
      "start": 288.5,
      "end": 290.34,
      "text": "后来我就参考它的idea。",
      "tokens": [
        51718,
        13547,
        6912,
        22020,
        2129,
        224,
        26504,
        45224,
        482,
        64,
        1543,
        51810
      ],
      "temperature": 0,
      "avg_logprob": -0.17641683058305221,
      "compression_ratio": 1.2290076335877862,
      "no_speech_prob": 1.254823155105722e-11,
      "source": "whisper"
    },
    {
      "id": 34,
      "seek": 29034,
      "start": 290.34,
      "end": 295.26,
      "text": "那现在让Deepseek V4 Flash来做这样的一个互动。",
      "tokens": [
        50365,
        4184,
        25040,
        33650,
        11089,
        595,
        405,
        916,
        691,
        19,
        20232,
        6912,
        10907,
        21209,
        1546,
        20182,
        1369,
        240,
        34961,
        1543,
        50611
      ],
      "temperature": 0,
      "avg_logprob": -0.2157778174190198,
      "compression_ratio": 0.9935897435897436,
      "no_speech_prob": 1.9529039843591e-11,
      "source": "whisper"
    },
    {
      "id": 35,
      "seek": 29034,
      "start": 296.32,
      "end": 301.52,
      "text": "点击画面里的不同元素是可以有互动的。",
      "tokens": [
        50664,
        12579,
        6336,
        119,
        27126,
        8833,
        15759,
        1546,
        47123,
        14812,
        37204,
        1541,
        6723,
        2412,
        1369,
        240,
        34961,
        1546,
        1543,
        50924
      ],
      "temperature": 0,
      "avg_logprob": -0.2157778174190198,
      "compression_ratio": 0.9935897435897436,
      "no_speech_prob": 1.9529039843591e-11,
      "source": "whisper"
    },
    {
      "id": 36,
      "seek": 29034,
      "start": 301.76,
      "end": 304.21999999999997,
      "text": "比如说刚刚这里窗户外有下雨。",
      "tokens": [
        50936,
        36757,
        8090,
        49160,
        49160,
        35102,
        30771,
        245,
        1486,
        115,
        12022,
        2412,
        4438,
        35339,
        1543,
        51059
      ],
      "temperature": 0,
      "avg_logprob": -0.2157778174190198,
      "compression_ratio": 0.9935897435897436,
      "no_speech_prob": 1.9529039843591e-11,
      "source": "whisper"
    },
    {
      "id": 37,
      "seek": 30422,
      "start": 304.22,
      "end": 308.72,
      "text": "然后这里点击咖啡豆呢,咖啡豆有溢出。",
      "tokens": [
        50365,
        26636,
        35102,
        12579,
        6336,
        119,
        8975,
        244,
        3284,
        94,
        45169,
        6240,
        11,
        8975,
        244,
        3284,
        94,
        45169,
        2412,
        16084,
        95,
        7781,
        1543,
        50590
      ],
      "temperature": 0,
      "avg_logprob": -0.12984900571862046,
      "compression_ratio": 1.3459119496855345,
      "no_speech_prob": 2.1774058250878525e-11,
      "source": "whisper"
    },
    {
      "id": 38,
      "seek": 30422,
      "start": 310.0,
      "end": 316.42,
      "text": "咖啡机点击之后,它会有蒸腾沸腾的效果。",
      "tokens": [
        50654,
        8975,
        244,
        3284,
        94,
        37960,
        12579,
        6336,
        119,
        9574,
        13547,
        11,
        11284,
        12949,
        2412,
        42356,
        116,
        21184,
        122,
        3308,
        116,
        21184,
        122,
        1546,
        43076,
        9319,
        1543,
        50975
      ],
      "temperature": 0,
      "avg_logprob": -0.12984900571862046,
      "compression_ratio": 1.3459119496855345,
      "no_speech_prob": 2.1774058250878525e-11,
      "source": "whisper"
    },
    {
      "id": 39,
      "seek": 30422,
      "start": 317.32000000000005,
      "end": 321.44000000000005,
      "text": "那点击人物,它这里会有插杯子的动作。",
      "tokens": [
        51020,
        4184,
        12579,
        6336,
        119,
        4035,
        23516,
        11,
        11284,
        35102,
        12949,
        2412,
        11673,
        240,
        45458,
        7626,
        1546,
        34961,
        11914,
        1543,
        51226
      ],
      "temperature": 0,
      "avg_logprob": -0.12984900571862046,
      "compression_ratio": 1.3459119496855345,
      "no_speech_prob": 2.1774058250878525e-11,
      "source": "whisper"
    },
    {
      "id": 40,
      "seek": 30422,
      "start": 321.78000000000003,
      "end": 324.54,
      "text": "点击咖啡杯呢,它也会做一个拉花的效果。",
      "tokens": [
        51243,
        12579,
        6336,
        119,
        8975,
        244,
        3284,
        94,
        45458,
        6240,
        11,
        11284,
        6404,
        12949,
        10907,
        20182,
        29717,
        20127,
        1546,
        43076,
        9319,
        1543,
        51381
      ],
      "temperature": 0,
      "avg_logprob": -0.12984900571862046,
      "compression_ratio": 1.3459119496855345,
      "no_speech_prob": 2.1774058250878525e-11,
      "source": "whisper"
    },
    {
      "id": 41,
      "seek": 32454,
      "start": 324.54,
      "end": 327.46000000000004,
      "text": "菜单也可以交互。",
      "tokens": [
        50365,
        33025,
        47446,
        47935,
        28455,
        1369,
        240,
        1543,
        50511
      ],
      "temperature": 0,
      "avg_logprob": -0.1697759159275743,
      "compression_ratio": 1.167883211678832,
      "no_speech_prob": 2.0007608417249578e-11,
      "source": "whisper"
    },
    {
      "id": 42,
      "seek": 32454,
      "start": 328.46000000000004,
      "end": 329.86,
      "text": "然后这里也有一些植物。",
      "tokens": [
        50561,
        26636,
        35102,
        6404,
        32241,
        13824,
        38739,
        235,
        23516,
        1543,
        50631
      ],
      "temperature": 0,
      "avg_logprob": -0.1697759159275743,
      "compression_ratio": 1.167883211678832,
      "no_speech_prob": 2.0007608417249578e-11,
      "source": "whisper"
    },
    {
      "id": 43,
      "seek": 32454,
      "start": 330.68,
      "end": 333.6,
      "text": "包括人物,这里的椅子也是可以移动。",
      "tokens": [
        50672,
        41828,
        4035,
        23516,
        11,
        35102,
        1546,
        38739,
        227,
        7626,
        22021,
        6723,
        8204,
        119,
        34961,
        1543,
        50818
      ],
      "temperature": 0,
      "avg_logprob": -0.1697759159275743,
      "compression_ratio": 1.167883211678832,
      "no_speech_prob": 2.0007608417249578e-11,
      "source": "whisper"
    },
    {
      "id": 44,
      "seek": 32454,
      "start": 334.82,
      "end": 337.1,
      "text": "桌子也是可以摇动。",
      "tokens": [
        50879,
        162,
        5475,
        7626,
        22021,
        6723,
        34783,
        229,
        34961,
        1543,
        50993
      ],
      "temperature": 0,
      "avg_logprob": -0.1697759159275743,
      "compression_ratio": 1.167883211678832,
      "no_speech_prob": 2.0007608417249578e-11,
      "source": "whisper"
    },
    {
      "id": 45,
      "seek": 32454,
      "start": 337.42,
      "end": 338.78000000000003,
      "text": "整体我感觉还不错。",
      "tokens": [
        51009,
        27662,
        29485,
        1654,
        9709,
        24447,
        14852,
        1960,
        29900,
        1543,
        51077
      ],
      "temperature": 0,
      "avg_logprob": -0.1697759159275743,
      "compression_ratio": 1.167883211678832,
      "no_speech_prob": 2.0007608417249578e-11,
      "source": "whisper"
    },
    {
      "id": 46,
      "seek": 33878,
      "start": 338.78,
      "end": 341.67999999999995,
      "text": "再给大家看一下我用OPS5生成的效果。",
      "tokens": [
        50365,
        8623,
        23197,
        6868,
        28324,
        1654,
        9254,
        46,
        6273,
        20,
        8244,
        11336,
        1546,
        43076,
        9319,
        1543,
        50510
      ],
      "temperature": 0,
      "avg_logprob": -0.33414306138691147,
      "compression_ratio": 0.8166666666666667,
      "no_speech_prob": 1.598808412051067e-11,
      "source": "whisper"
    },
    {
      "id": 47,
      "seek": 34168,
      "start": 341.68,
      "end": 342.68,
      "text": "。",
      "tokens": [
        50365,
        1543,
        50415
      ],
      "temperature": 0,
      "avg_logprob": -1.316196084022522,
      "compression_ratio": 0.2727272727272727,
      "no_speech_prob": 2.842077406750576e-10,
      "source": "whisper"
    },
    {
      "id": 48,
      "seek": 37168,
      "start": 371.68,
      "end": 372.68,
      "text": "。",
      "tokens": [
        50365,
        1543,
        50415
      ],
      "temperature": 0,
      "avg_logprob": -0.7986320495605469,
      "compression_ratio": 1.7142857142857142,
      "no_speech_prob": 3.8750944120202746e-10,
      "source": "whisper"
    },
    {
      "id": 51,
      "seek": 37168,
      "start": 372.68,
      "end": 373.68,
      "text": "。",
      "tokens": [
        50415,
        1543,
        50465
      ],
      "temperature": 0,
      "avg_logprob": -0.7986320495605469,
      "compression_ratio": 1.7142857142857142,
      "no_speech_prob": 3.8750944120202746e-10,
      "source": "whisper"
    },
    {
      "id": 56,
      "seek": 40168,
      "start": 401.68,
      "end": 404.62,
      "text": "OPS5呢,它是我提示了它两次。",
      "tokens": [
        50365,
        46,
        6273,
        20,
        6240,
        11,
        11284,
        42614,
        20949,
        25696,
        2289,
        11284,
        36257,
        9487,
        1543,
        50512
      ],
      "temperature": 0,
      "avg_logprob": -0.09440934922960069,
      "compression_ratio": 1.2244897959183674,
      "no_speech_prob": 1.4745243220870563e-11,
      "source": "whisper"
    },
    {
      "id": 57,
      "seek": 40168,
      "start": 405.22,
      "end": 410.32,
      "text": "它一开始呢,对画面里做的交互不是很丰富。",
      "tokens": [
        50542,
        11284,
        2257,
        45213,
        6240,
        11,
        8713,
        27126,
        8833,
        15759,
        10907,
        1546,
        28455,
        1369,
        240,
        7296,
        4563,
        940,
        108,
        47564,
        1543,
        50797
      ],
      "temperature": 0,
      "avg_logprob": -0.09440934922960069,
      "compression_ratio": 1.2244897959183674,
      "no_speech_prob": 1.4745243220870563e-11,
      "source": "whisper"
    },
    {
      "id": 58,
      "seek": 40168,
      "start": 410.8,
      "end": 413.42,
      "text": "所以后来我又提示让它去优化一下。",
      "tokens": [
        50821,
        7239,
        13547,
        6912,
        1654,
        17047,
        20949,
        25696,
        33650,
        11284,
        6734,
        7384,
        246,
        23756,
        8861,
        1543,
        50952
      ],
      "temperature": 0,
      "avg_logprob": -0.09440934922960069,
      "compression_ratio": 1.2244897959183674,
      "no_speech_prob": 1.4745243220870563e-11,
      "source": "whisper"
    },
    {
      "id": 59,
      "seek": 40168,
      "start": 413.84000000000003,
      "end": 415.90000000000003,
      "text": "而且它的整体速度是非常慢的。",
      "tokens": [
        50973,
        22942,
        45224,
        27662,
        29485,
        31217,
        13127,
        1541,
        14392,
        20645,
        1546,
        1543,
        51076
      ],
      "temperature": 0,
      "avg_logprob": -0.09440934922960069,
      "compression_ratio": 1.2244897959183674,
      "no_speech_prob": 1.4745243220870563e-11,
      "source": "whisper"
    },
    {
      "id": 60,
      "seek": 40168,
      "start": 416.34000000000003,
      "end": 421.7,
      "text": "当然它花费的价格也是远远高于V4 Flash的。",
      "tokens": [
        51098,
        40486,
        11284,
        20127,
        18464,
        117,
        1546,
        1550,
        115,
        30921,
        22021,
        3316,
        250,
        3316,
        250,
        12979,
        37732,
        53,
        19,
        20232,
        1546,
        1543,
        51366
      ],
      "temperature": 0,
      "avg_logprob": -0.09440934922960069,
      "compression_ratio": 1.2244897959183674,
      "no_speech_prob": 1.4745243220870563e-11,
      "source": "whisper"
    },
    {
      "id": 61,
      "seek": 40168,
      "start": 422.06,
      "end": 425.4,
      "text": "其实我还有两个任务呢,让V4 Flash在运行着。",
      "tokens": [
        51384,
        41646,
        1654,
        35091,
        36257,
        7549,
        26443,
        5087,
        94,
        6240,
        11,
        33650,
        53,
        19,
        20232,
        3581,
        3316,
        238,
        8082,
        20708,
        1543,
        51551
      ],
      "temperature": 0,
      "avg_logprob": -0.09440934922960069,
      "compression_ratio": 1.2244897959183674,
      "no_speech_prob": 1.4745243220870563e-11,
      "source": "whisper"
    },
    {
      "id": 62,
      "seek": 40168,
      "start": 426.0,
      "end": 428.96000000000004,
      "text": "那它们已经运行了一个多小时了,还没有结束。",
      "tokens": [
        51581,
        4184,
        11284,
        9497,
        49161,
        3316,
        238,
        46606,
        20182,
        6392,
        7322,
        15729,
        2289,
        11,
        14852,
        17944,
        45641,
        44320,
        1543,
        51729
      ],
      "temperature": 0,
      "avg_logprob": -0.09440934922960069,
      "compression_ratio": 1.2244897959183674,
      "no_speech_prob": 1.4745243220870563e-11,
      "source": "whisper"
    },
    {
      "id": 63,
      "seek": 42896,
      "start": 428.96,
      "end": 432.23999999999995,
      "text": "其实我在日常里呢,还是会经常用到V4 Flash的。",
      "tokens": [
        50365,
        41646,
        40528,
        6890,
        11279,
        15759,
        6240,
        11,
        45726,
        12949,
        30276,
        11279,
        9254,
        4511,
        53,
        19,
        20232,
        1546,
        1543,
        50529
      ],
      "temperature": 0,
      "avg_logprob": -0.18043174473106438,
      "compression_ratio": 1.194267515923567,
      "no_speech_prob": 1.4212016979930908e-11,
      "source": "whisper"
    },
    {
      "id": 64,
      "seek": 42896,
      "start": 432.62,
      "end": 434.29999999999995,
      "text": "比如说现在看到的这个应用呢,",
      "tokens": [
        50548,
        36757,
        8090,
        25040,
        18032,
        1546,
        15368,
        44297,
        9254,
        6240,
        11,
        50632
      ],
      "temperature": 0,
      "avg_logprob": -0.18043174473106438,
      "compression_ratio": 1.194267515923567,
      "no_speech_prob": 1.4212016979930908e-11,
      "source": "whisper"
    },
    {
      "id": 65,
      "seek": 42896,
      "start": 434.97999999999996,
      "end": 439.26,
      "text": "我就是在半小时之内,让Glock 4.5生成的。",
      "tokens": [
        50666,
        1654,
        5620,
        3581,
        30018,
        7322,
        15729,
        9574,
        34742,
        11,
        33650,
        38,
        4102,
        1017,
        13,
        20,
        8244,
        11336,
        1546,
        1543,
        50880
      ],
      "temperature": 0,
      "avg_logprob": -0.18043174473106438,
      "compression_ratio": 1.194267515923567,
      "no_speech_prob": 1.4212016979930908e-11,
      "source": "whisper"
    },
    {
      "id": 66,
      "seek": 42896,
      "start": 439.91999999999996,
      "end": 443.02,
      "text": "揭露的是Dipsic V4 Flash和Pro模型。",
      "tokens": [
        50913,
        11673,
        255,
        18594,
        110,
        24620,
        35,
        2600,
        299,
        691,
        19,
        20232,
        12565,
        12681,
        41908,
        39823,
        1543,
        51068
      ],
      "temperature": 0,
      "avg_logprob": -0.18043174473106438,
      "compression_ratio": 1.194267515923567,
      "no_speech_prob": 1.4212016979930908e-11,
      "source": "whisper"
    },
    {
      "id": 67,
      "seek": 42896,
      "start": 443.02,
      "end": 446.32,
      "text": "因为它本身是EM的上下文窗口。",
      "tokens": [
        51068,
        34627,
        11284,
        8802,
        19847,
        1541,
        6683,
        1546,
        5708,
        4438,
        17174,
        30771,
        245,
        18144,
        1543,
        51233
      ],
      "temperature": 0,
      "avg_logprob": -0.18043174473106438,
      "compression_ratio": 1.194267515923567,
      "no_speech_prob": 1.4212016979930908e-11,
      "source": "whisper"
    },
    {
      "id": 68,
      "seek": 42896,
      "start": 446.76,
      "end": 449.74,
      "text": "而且它的生成速度很快,价格很便宜。",
      "tokens": [
        51255,
        22942,
        45224,
        8244,
        11336,
        31217,
        13127,
        4563,
        10251,
        11,
        1550,
        115,
        30921,
        4563,
        27364,
        2415,
        250,
        1543,
        51404
      ],
      "temperature": 0,
      "avg_logprob": -0.18043174473106438,
      "compression_ratio": 1.194267515923567,
      "no_speech_prob": 1.4212016979930908e-11,
      "source": "whisper"
    },
    {
      "id": 69,
      "seek": 42896,
      "start": 450.26,
      "end": 454.96,
      "text": "所以我会把它揭露在我自己用的一些App里,",
      "tokens": [
        51430,
        7239,
        1654,
        12949,
        42061,
        11673,
        255,
        18594,
        110,
        3581,
        1654,
        17645,
        9254,
        1546,
        38515,
        9132,
        15759,
        11,
        51665
      ],
      "temperature": 0,
      "avg_logprob": -0.18043174473106438,
      "compression_ratio": 1.194267515923567,
      "no_speech_prob": 1.4212016979930908e-11,
      "source": "whisper"
    },
    {
      "id": 70,
      "seek": 42896,
      "start": 455.5,
      "end": 458.4,
      "text": "或者是用于翻译的插件里。",
      "tokens": [
        51692,
        31148,
        1541,
        9254,
        37732,
        42716,
        5233,
        239,
        1546,
        11673,
        240,
        20485,
        15759,
        1543,
        51837
      ],
      "temperature": 0,
      "avg_logprob": -0.18043174473106438,
      "compression_ratio": 1.194267515923567,
      "no_speech_prob": 1.4212016979930908e-11,
      "source": "whisper"
    },
    {
      "id": 71,
      "seek": 45896,
      "start": 458.96,
      "end": 462.46,
      "text": "都是速度、质量都是非常好的。",
      "tokens": [
        50365,
        22796,
        31217,
        13127,
        1231,
        18464,
        101,
        26748,
        22796,
        14392,
        20715,
        1543,
        50540
      ],
      "temperature": 0,
      "avg_logprob": -0.19078520572546756,
      "compression_ratio": 1.1071428571428572,
      "no_speech_prob": 1.3034863986793876e-11,
      "source": "whisper"
    },
    {
      "id": 72,
      "seek": 45896,
      "start": 462.73999999999995,
      "end": 465.08,
      "text": "V4 Flash的应用场景还是非常多的。",
      "tokens": [
        50554,
        53,
        19,
        20232,
        1546,
        44297,
        9254,
        50255,
        50218,
        45726,
        14392,
        6392,
        1546,
        1543,
        50671
      ],
      "temperature": 0,
      "avg_logprob": -0.19078520572546756,
      "compression_ratio": 1.1071428571428572,
      "no_speech_prob": 1.3034863986793876e-11,
      "source": "whisper"
    },
    {
      "id": 73,
      "seek": 45896,
      "start": 465.29999999999995,
      "end": 469.64,
      "text": "但就编程而言,因为我们现在也是有大量的,",
      "tokens": [
        50682,
        8395,
        3111,
        38109,
        244,
        29649,
        11070,
        12009,
        11,
        34627,
        15003,
        25040,
        22021,
        2412,
        3582,
        26748,
        1546,
        11,
        50899
      ],
      "temperature": 0,
      "avg_logprob": -0.19078520572546756,
      "compression_ratio": 1.1071428571428572,
      "no_speech_prob": 1.3034863986793876e-11,
      "source": "whisper"
    },
    {
      "id": 74,
      "seek": 45896,
      "start": 469.91999999999996,
      "end": 472.32,
      "text": "更大规模的前沿模型可以使用。",
      "tokens": [
        50913,
        19002,
        3582,
        6758,
        226,
        41908,
        1546,
        8945,
        3308,
        123,
        41908,
        39823,
        6723,
        22982,
        9254,
        1543,
        51033
      ],
      "temperature": 0,
      "avg_logprob": -0.19078520572546756,
      "compression_ratio": 1.1071428571428572,
      "no_speech_prob": 1.3034863986793876e-11,
      "source": "whisper"
    },
    {
      "id": 75,
      "seek": 45896,
      "start": 472.41999999999996,
      "end": 476.15999999999997,
      "text": "无论是K3,Office,或者是GPT-Sore模型,",
      "tokens": [
        51038,
        31364,
        7422,
        118,
        1541,
        42,
        18,
        11,
        29745,
        573,
        11,
        31148,
        1541,
        38,
        47,
        51,
        12,
        50,
        418,
        41908,
        39823,
        11,
        51225
      ],
      "temperature": 0,
      "avg_logprob": -0.19078520572546756,
      "compression_ratio": 1.1071428571428572,
      "no_speech_prob": 1.3034863986793876e-11,
      "source": "whisper"
    },
    {
      "id": 76,
      "seek": 45896,
      "start": 476.29999999999995,
      "end": 477.71999999999997,
      "text": "或者是Glock 4.5。",
      "tokens": [
        51232,
        31148,
        1541,
        38,
        4102,
        1017,
        13,
        20,
        1543,
        51303
      ],
      "temperature": 0,
      "avg_logprob": -0.19078520572546756,
      "compression_ratio": 1.1071428571428572,
      "no_speech_prob": 1.3034863986793876e-11,
      "source": "whisper"
    },
    {
      "id": 77,
      "seek": 47772,
      "start": 477.72,
      "end": 479.44000000000005,
      "text": "那我个人认为的话,",
      "tokens": [
        50365,
        4184,
        1654,
        7549,
        4035,
        7422,
        97,
        13992,
        44575,
        11,
        50451
      ],
      "temperature": 0,
      "avg_logprob": -0.22022153893295598,
      "compression_ratio": 1.0177777777777777,
      "no_speech_prob": 1.163281363597557e-11,
      "source": "whisper"
    },
    {
      "id": 78,
      "seek": 47772,
      "start": 479.94000000000005,
      "end": 483.52000000000004,
      "text": "V4 Flash正式版在编码上还是弱了一些。",
      "tokens": [
        50476,
        53,
        19,
        20232,
        15789,
        27584,
        42096,
        3581,
        38109,
        244,
        23230,
        223,
        5708,
        45726,
        5702,
        109,
        2289,
        38515,
        1543,
        50655
      ],
      "temperature": 0,
      "avg_logprob": -0.22022153893295598,
      "compression_ratio": 1.0177777777777777,
      "no_speech_prob": 1.163281363597557e-11,
      "source": "whisper"
    },
    {
      "id": 79,
      "seek": 47772,
      "start": 483.8,
      "end": 485.0,
      "text": "简单的,我让它执行OK,",
      "tokens": [
        50669,
        11249,
        222,
        47446,
        1546,
        11,
        1654,
        33650,
        11284,
        3416,
        100,
        8082,
        9443,
        11,
        50729
      ],
      "temperature": 0,
      "avg_logprob": -0.22022153893295598,
      "compression_ratio": 1.0177777777777777,
      "no_speech_prob": 1.163281363597557e-11,
      "source": "whisper"
    },
    {
      "id": 80,
      "seek": 47772,
      "start": 485.26000000000005,
      "end": 486.42,
      "text": "但是更复杂的。",
      "tokens": [
        50742,
        11189,
        19002,
        1787,
        235,
        4422,
        224,
        1546,
        1543,
        50800
      ],
      "temperature": 0,
      "avg_logprob": -0.22022153893295598,
      "compression_ratio": 1.0177777777777777,
      "no_speech_prob": 1.163281363597557e-11,
      "source": "whisper"
    },
    {
      "id": 81,
      "seek": 47772,
      "start": 487.06,
      "end": 492.14000000000004,
      "text": "或者说你更想OneShot来去生成一个比较好的结果的话,",
      "tokens": [
        50832,
        31148,
        8090,
        2166,
        19002,
        7093,
        15426,
        7774,
        310,
        6912,
        6734,
        8244,
        11336,
        20182,
        11706,
        9830,
        225,
        20715,
        45641,
        9319,
        44575,
        11,
        51086
      ],
      "temperature": 0,
      "avg_logprob": -0.22022153893295598,
      "compression_ratio": 1.0177777777777777,
      "no_speech_prob": 1.163281363597557e-11,
      "source": "whisper"
    },
    {
      "id": 82,
      "seek": 47772,
      "start": 492.90000000000003,
      "end": 496.92,
      "text": "可能我会更倾向于用GPT模型,",
      "tokens": [
        51124,
        16657,
        1654,
        12949,
        19002,
        2122,
        122,
        24282,
        37732,
        9254,
        38,
        47,
        51,
        41908,
        39823,
        11,
        51325
      ],
      "temperature": 0,
      "avg_logprob": -0.22022153893295598,
      "compression_ratio": 1.0177777777777777,
      "no_speech_prob": 1.163281363597557e-11,
      "source": "whisper"
    },
    {
      "id": 83,
      "seek": 49692,
      "start": 496.92,
      "end": 498.38,
      "text": "或者是Glock模型。",
      "tokens": [
        50365,
        31148,
        1541,
        38,
        4102,
        41908,
        39823,
        1543,
        50438
      ],
      "temperature": 0,
      "avg_logprob": -0.23558921284145778,
      "compression_ratio": 0.9105691056910569,
      "no_speech_prob": 1.140748693423399e-11,
      "source": "whisper"
    },
    {
      "id": 84,
      "seek": 49692,
      "start": 498.62,
      "end": 500.54,
      "text": "以上就是今天的所有分享,",
      "tokens": [
        50450,
        29497,
        5620,
        34947,
        39300,
        30855,
        11,
        50546
      ],
      "temperature": 0,
      "avg_logprob": -0.23558921284145778,
      "compression_ratio": 0.9105691056910569,
      "no_speech_prob": 1.140748693423399e-11,
      "source": "whisper"
    },
    {
      "id": 85,
      "seek": 49692,
      "start": 501.2,
      "end": 502.64000000000004,
      "text": "希望我的视频对你有帮助,",
      "tokens": [
        50579,
        29955,
        14200,
        40656,
        39752,
        8713,
        43320,
        4845,
        106,
        37618,
        11,
        50651
      ],
      "temperature": 0,
      "avg_logprob": -0.23558921284145778,
      "compression_ratio": 0.9105691056910569,
      "no_speech_prob": 1.140748693423399e-11,
      "source": "whisper"
    },
    {
      "id": 86,
      "seek": 49692,
      "start": 502.90000000000003,
      "end": 503.62,
      "text": "我们下次再见。",
      "tokens": [
        50664,
        15003,
        35829,
        44176,
        1543,
        50700
      ],
      "temperature": 0,
      "avg_logprob": -0.23558921284145778,
      "compression_ratio": 0.9105691056910569,
      "no_speech_prob": 1.140748693423399e-11,
      "source": "whisper"
    }
  ],
  "transcript_selection": {
    "schema": "videonote.transcript_selection.v1",
    "status": "SELECTED",
    "requested_language": "zh",
    "duration": 503.965938,
    "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": 81,
      "first_start": 0.0,
      "final_end": 503.62,
      "duration_coverage": 0.999314,
      "maximum_gap_seconds": 29.0,
      "inverted_count": 0,
      "language": {
        "requested": "zh",
        "matched": true,
        "latin": 343,
        "han": 1515,
        "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": []
    }
  }
}