1
00:00:00,000 --> 00:00:02,320
GPT-6Astra is here.

2
00:00:02,320 --> 00:00:04,920
Today we are releasing Fable 5.1.

3
00:00:04,920 --> 00:00:09,120
Your action is to absorb one or two atomic pieces of information.

4
00:00:09,120 --> 00:00:13,920
There is absolutely no need to completely overturn and scrap your original workflow.

5
00:00:13,920 --> 00:00:16,720
You actually only have three choices at hand.

6
00:00:16,720 --> 00:00:19,200
You only have three choices at hand.

7
00:00:19,200 --> 00:00:22,080
CloudOps 5.5 Video Editing.

8
00:00:22,080 --> 00:00:24,640
Is it really as good as people say?

9
00:00:24,640 --> 00:00:29,520
The short clips that flashed on screen were all edited by it entirely on its own.

10
00:00:29,520 --> 00:00:31,920
我沒有在意義中在這裏的程度

11
00:00:31,920 --> 00:00:33,120
它已經發出了

12
00:00:33,120 --> 00:00:36,720
然後在這兩天的電腦都在發出了

13
00:00:36,720 --> 00:00:39,120
一些人用它們甚至用它們的一段時間

14
00:00:39,120 --> 00:00:42,120
我沒有只是看過 demos picked by others

15
00:00:42,120 --> 00:00:43,920
所以我把這裏的十分時間

16
00:00:43,920 --> 00:00:45,720
十分時間的電話題

17
00:00:45,720 --> 00:00:47,320
我的樣子和它們都做了

18
00:00:47,320 --> 00:00:50,520
和它們做了整個作品的工作

19
00:00:50,520 --> 00:00:53,600
這點是,它們的確是強烈

20
00:00:53,600 --> 00:00:54,720
比我還想的

21
00:00:54,720 --> 00:00:56,520
尤其是在其中的地方

22
00:00:56,520 --> 00:00:58,720
我認為它們是完全不可怕的

23
00:00:58,720 --> 00:01:02,020
而是在做什麼工作是在做的,

24
00:01:02,020 --> 00:01:05,120
在看這個影片,你可能會有更多想法。

25
00:01:05,120 --> 00:01:07,720
當然,你可能會更多疑惑。

26
00:01:07,720 --> 00:01:10,520
Hey! Hello! And welcome back to Channel Name!

27
00:01:10,520 --> 00:01:13,520
The conversation between the two conferences in the intro

28
00:01:13,520 --> 00:01:16,420
was actually just an appetizer I made on a whim.

29
00:01:16,420 --> 00:01:19,220
Both companies released their own conference videos,

30
00:01:19,220 --> 00:01:22,420
one over three minutes, and the other a minute and a half,

31
00:01:22,420 --> 00:01:24,220
both with English subtitles,

32
00:01:24,220 --> 00:01:28,220
so I had the model edit these two clips into a mixed dialogue scene,

33
00:01:28,220 --> 00:01:30,220
and add bilingual subtitles.

34
00:01:30,220 --> 00:01:32,560
This was actually quite troublesome at first,

35
00:01:32,560 --> 00:01:35,320
because the English subtitles included with the two videos

36
00:01:35,320 --> 00:01:37,220
were scrolling auto captions,

37
00:01:37,220 --> 00:01:39,160
and the timestamps were all overlapping,

38
00:01:39,160 --> 00:01:41,400
making it impossible to determine the cut points.

39
00:01:41,400 --> 00:01:43,740
After CloudOps 5.5 detected this issue,

40
00:01:43,740 --> 00:01:45,040
it called Whisper itself,

41
00:01:45,040 --> 00:01:46,440
retranscribed everything,

42
00:01:46,440 --> 00:01:48,840
and only then was it able to accurately sync the cuts.

43
00:01:48,840 --> 00:01:50,800
The final clip is 66 seconds long,

44
00:01:50,800 --> 00:01:53,240
and the result is a bit beyond my expectations.

45
00:01:53,240 --> 00:01:55,080
GPT6 Astra is here.

46
00:01:55,080 --> 00:01:56,740
It's our latest frontier model,

47
00:01:56,740 --> 00:02:00,860
and the best model in the world for tasks where raw intelligence matters.

48
00:02:00,860 --> 00:02:05,980
Today we're releasing Fable 5.1, the latest upgrade to our most capable model class.

49
00:02:05,980 --> 00:02:10,500
I'm able to hand it bigger, less well-defined tasks with minimal hand-holding.

50
00:02:10,500 --> 00:02:15,340
Whatever you already handle, it can take on more and do the hardest parts better.

51
00:02:15,340 --> 00:02:18,580
And whether you're coding, writing, designing,

52
00:02:18,580 --> 00:02:22,260
or just working on things that require a little bit more intelligence,

53
00:02:22,260 --> 00:02:26,380
you'll find that Astra delivers great results, often on the first attempt.

54
00:02:26,380 --> 00:02:30,540
如果它 hits a wall,它 tells you what它 tried and where it got stuck.

55
00:02:30,540 --> 00:02:34,460
As you might expect, Astra excels at long-running tasks.

56
00:02:34,460 --> 00:02:39,740
These are the kinds of tasks where a small mistake in Step 2 messes things up in Step 40,

57
00:02:39,740 --> 00:02:42,460
and Fable 5.1 holds up the whole way.

58
00:02:42,460 --> 00:02:45,980
The model is able to paint a hopefully recognizable picture

59
00:02:45,980 --> 00:02:48,780
far faster and far better than I could have done.

60
00:02:48,780 --> 00:02:51,500
And this model is great at science.

61
00:02:51,500 --> 00:02:52,460
Try it on!

62
00:02:56,380 --> 00:03:03,100
我以前想要放在同一篇的畫面

63
00:03:03,100 --> 00:03:06,060
所以他們看起來像是在同一場場

64
00:03:06,060 --> 00:03:11,340
但我认为这个分别的课程是很好的。

65
00:03:11,340 --> 00:03:13,820
没有必要在手机上的手机上的手机上,

66
00:03:13,820 --> 00:03:16,780
它们的手机上会在纳续上的角色,

67
00:03:16,780 --> 00:03:18,780
而其他手机上会在几个空间的手机上,

68
00:03:18,780 --> 00:03:21,660
让整个整个话会很滑。

69
00:03:21,660 --> 00:03:24,780
这些字幕上是很很平均的,

70
00:03:24,780 --> 00:03:25,980
而且是有趣的,

71
00:03:25,980 --> 00:03:29,660
它甚至是一种小细节的细节上的纳细节。

72
00:03:29,660 --> 00:03:30,700
《MATERIALS》

73
00:03:30,700 --> 00:03:34,460
《Spliced Clips》从两个公司的公司公司公司公司的影片。

74
00:03:34,460 --> 00:03:36,620
我們可以甚至出一個架頭的架頭

75
00:03:36,620 --> 00:03:40,700
說兩者公司公司公司推出了強烈的模特兒

76
00:03:40,700 --> 00:03:43,500
當然,這座的公開是一種熱烈的表現

77
00:03:43,500 --> 00:03:48,300
今天真正的挑戰是我的17-27-second網上網上

78
00:03:48,300 --> 00:03:52,460
你會想到會有很多網上網上看過很多網上的示範

79
00:03:52,460 --> 00:03:55,260
是它是隱屬的充滿的 plug-ins

80
00:03:55,260 --> 00:03:57,340
或者千萬多文字和文字的文字

81
00:03:57,340 --> 00:03:59,100
就能達成這個樣子嗎

82
00:03:59,100 --> 00:04:02,300
這次,它完全是完全手上

83
00:04:02,300 --> 00:04:05,860
I gave it a pristine cloud code command line environment

84
00:04:05,860 --> 00:04:09,500
with no custom skills loaded, no specialized rules written,

85
00:04:09,500 --> 00:04:12,620
no MCP, and no preset editing workflows.

86
00:04:12,620 --> 00:04:15,060
There were only two folders in the working directory,

87
00:04:15,060 --> 00:04:16,340
one for the raw footage,

88
00:04:16,340 --> 00:04:19,620
and one empty output directory for transcriptions and clips.

89
00:04:19,620 --> 00:04:23,260
It handled all this work itself by calling local tools,

90
00:04:23,260 --> 00:04:25,060
without anyone needing to teach it.

91
00:04:25,060 --> 00:04:28,500
Seven processes corresponded to seven prompts.

92
00:04:28,500 --> 00:04:31,980
Once I handed them over, it ran fully automatically.

93
00:04:31,980 --> 00:04:33,980
我沒有按一指,

94
00:05:33,980 --> 00:05:35,980
在影片 description中,

95
00:05:35,980 --> 00:05:37,980
它先叫一按照

96
00:06:07,980 --> 00:06:09,980
這就是同樣的同樣的同樣的同樣

97
00:06:09,980 --> 00:06:11,980
原來是97秒

98
00:06:11,980 --> 00:06:13,980
然後再拍攝,它是83.7秒

99
00:06:13,980 --> 00:06:15,980
我們看看這段時間的故事

100
00:06:15,980 --> 00:06:17,980
我先說過這次的情況

101
00:06:17,980 --> 00:06:19,980
我再說過這次的情況

102
00:06:19,980 --> 00:06:21,980
我再說過這次的情況

103
00:06:21,980 --> 00:06:23,980
它們的確定了

104
00:06:53,980 --> 00:06:56,980
你其實只有三個選擇

105
00:06:56,980 --> 00:07:00,980
第一次選擇

106
00:07:00,980 --> 00:07:02,980
老問題,老計

107
00:07:02,980 --> 00:07:05,980
你其實只有三個選擇

108
00:07:05,980 --> 00:07:07,980
第一次選擇

109
00:07:07,980 --> 00:07:08,980
老問題,老計

110
00:07:08,980 --> 00:07:10,980
四十四十字

111
00:07:10,980 --> 00:07:11,980
沒有一名 typo

112
00:07:11,980 --> 00:07:13,980
nor did it change my original meaning

113
00:07:13,980 --> 00:07:15,980
有另一個有趣的原因

114
00:07:15,980 --> 00:07:17,980
在我中間說

115
00:07:17,980 --> 00:07:18,980
我一句說

116
00:07:18,980 --> 00:07:19,980
是解決的

117
00:07:19,980 --> 00:07:20,980
是一樣的

118
00:07:20,980 --> 00:07:21,980
是以前的

119
00:07:21,980 --> 00:07:27,220
这个模特别也提到这个文章也提到这个文章也说这可能会是一种语言的语言

120
00:07:27,220 --> 00:07:29,620
它不太清楚是否一名字,

121
00:07:29,620 --> 00:07:31,140
Solution,或Method,

122
00:07:31,140 --> 00:07:33,060
但它 eventually就算是否保持它。

123
00:07:33,060 --> 00:07:34,180
我聽到它,

124
00:07:34,180 --> 00:07:35,860
我認為它是正確的,

125
00:07:35,860 --> 00:07:38,580
這是一種自然的自然而言,

126
00:07:38,580 --> 00:07:39,620
不是一種 error。

127
00:07:39,620 --> 00:07:42,180
它主問題是解決的,

128
00:07:42,180 --> 00:07:43,060
而是解決的,

129
00:07:43,060 --> 00:07:45,140
而是解決的一樣。

130
00:07:45,140 --> 00:07:47,220
而且,有一個不足的 bonus。

131
00:07:47,220 --> 00:07:49,220
在916的影片,

132
00:07:49,220 --> 00:07:50,740
從1190的影片,

133
00:07:50,740 --> 00:07:52,180
它其實相反轉了

134
00:07:52,180 --> 00:07:54,020
中間和接觸的剪接

135
00:07:54,020 --> 00:07:54,740
在剪接的剪接,

136
00:07:54,740 --> 00:07:55,940
在剪接的剪接。

137
00:07:55,940 --> 00:07:59,940
在最終的作品,我的頭髮看起來更大,

138
00:07:59,940 --> 00:08:01,700
而且還不錯。

139
00:08:01,700 --> 00:08:04,180
這是像是推進和推出的照片

140
00:08:04,180 --> 00:08:07,060
很多影片使用來令觀眾失散了。

141
00:08:07,060 --> 00:08:09,700
如果是一件事,或是一件事,

142
00:08:09,700 --> 00:08:11,780
最後的效果是非常舒服。

143
00:08:11,780 --> 00:08:13,860
OK,一只短片是一整個短片,

144
00:08:13,860 --> 00:08:16,740
但什麼時候會逐測 multiple segments?

145
00:08:16,740 --> 00:08:20,180
第三次,我讓它選擇四個整個整個整個整個

146
00:08:20,180 --> 00:08:21,540
整個短片。

147
00:08:21,540 --> 00:08:27,840
它選了四個成員,從22秒,5秒23秒,8秒04秒,16秒27秒

148
00:08:27,840 --> 00:08:32,780
我的題目是名字《Don't Get Led by the Nose When New AI Terms Keep Popping Up》

149
00:08:32,780 --> 00:08:35,040
我將成員的成員作成為一個成員作成的一段時間

150
00:08:35,040 --> 00:08:40,200
因為原本已經形成了一段時間的成員作成為一個成員作成的成員作成員

151
00:08:40,200 --> 00:08:42,960
的問題, 理解, 理解, 理解, 理解

152
00:08:42,960 --> 00:08:47,000
它們的遙控和聲音和聲音和聲音

153
00:08:47,000 --> 00:08:49,640
遙控了些泡沫中的聲音

154
00:08:49,640 --> 00:08:52,040
和聲音和聲音和聲音

155
00:09:07,233 --> 00:09:09,233
你认为了解的讯息是有关的,

156
00:09:09,233 --> 00:09:12,233
你会感到的解释的解释。

157
00:09:12,233 --> 00:09:16,233
你会回到那终于学习的终于学习的终于学习的终于,

158
00:09:16,233 --> 00:09:19,233
但这些解释的解释是很简单的。

159
00:09:19,233 --> 00:09:23,233
只要是在其中的解释的解释,

160
00:09:23,233 --> 00:09:25,233
它根本的解释是没有改变。

161
00:09:25,233 --> 00:09:27,233
这就是最大的疾病,

162
00:09:27,233 --> 00:09:29,233
在整个集中的解释是最大的解释。

163
00:09:29,233 --> 00:09:31,233
当第三个解释开始,

164
00:09:31,233 --> 00:09:32,233
我说,

165
00:09:32,233 --> 00:09:36,233
只要是在其中的解释的解释,

166
00:09:36,233 --> 00:09:37,233
而是它根本的缺乏,

167
00:09:37,233 --> 00:09:39,433
它根本的缺乏没有改变。

168
00:09:39,433 --> 00:09:40,633
在这个词中,

169
00:09:40,633 --> 00:09:41,873
這技能

170
00:09:41,873 --> 00:09:44,873
在原本的影片,我剛才說了我的能力

171
00:09:44,873 --> 00:09:48,033
表現了,所以它流出了很容易

172
00:09:48,033 --> 00:09:50,873
但在這個78-second short他在一起

173
00:09:50,873 --> 00:09:54,633
沒有提及的名字,所以觀眾會被聽到

174
00:09:54,633 --> 00:09:57,033
所以觀眾會被聽到這次聽到

175
00:09:57,033 --> 00:09:58,633
為什麼會這個事?

176
00:09:58,633 --> 00:09:59,833
我查看他的報告,

177
00:09:59,833 --> 00:10:02,633
而 GPT-4.5其實看到了

178
00:10:02,633 --> 00:10:04,633
這個文字的文字是太 broad,

179
00:10:04,633 --> 00:10:06,633
但它覺得這個文字的文字

180
00:10:06,633 --> 00:10:09,633
在前面的前面的文字可以遮到它,

181
00:10:09,633 --> 00:10:11,633
所以它不適合它,

182
00:10:11,633 --> 00:10:16,393
但是,在人的耳耳耳,你能聽到一聲音的音效

183
00:10:16,393 --> 00:10:18,793
倘僅僅僅是在語音的語音

184
00:10:18,793 --> 00:10:21,233
不支持這個語音的語音

185
00:10:21,233 --> 00:10:22,833
這是一個問題的用力

186
00:10:22,833 --> 00:10:25,593
它切音的音效很清晰

187
00:10:25,593 --> 00:10:28,393
但因為它們的音效受到音效的音效

188
00:10:28,393 --> 00:10:31,273
它們不適合這個音效的音效

189
00:10:31,273 --> 00:10:35,393
所以,這步還要有用的手段

190
00:10:35,393 --> 00:10:37,073
我聽到整個動作

191
00:10:37,273 --> 00:10:38,793
我發現這個解釋

192
00:10:38,793 --> 00:10:42,173
and once I caught this disconnect, I gave him revision requirements.

193
00:10:42,173 --> 00:10:45,953
I pointed out that specific spot and set very strict constraints.

194
00:10:45,953 --> 00:10:46,553
I told him,

195
00:10:46,553 --> 00:10:48,853
the audience won't understand this ability,

196
00:10:48,853 --> 00:10:50,953
so you need to find a way to make the transition natural.

197
00:10:50,953 --> 00:10:54,373
You can rearrange sentences, swap paragraphs, or change the order,

198
00:10:54,373 --> 00:10:57,413
but you must absolutely not add anything I didn't say,

199
00:10:57,413 --> 00:10:58,793
nor can you distort my meaning.

200
00:10:58,793 --> 00:11:00,753
Don't touch a single thing anywhere else.

201
00:11:00,753 --> 00:11:02,453
Under these strict constraints,

202
00:11:02,453 --> 00:11:06,293
everyone's first reaction would be that it had to delete that sentence.

203
00:11:06,293 --> 00:11:07,913
That's the simplest way.

204
00:11:07,913 --> 00:11:11,913
但它完全改正了我的意見

205
00:11:11,913 --> 00:11:13,913
它不在意見,

206
00:11:13,913 --> 00:11:15,913
但它們不在意見,

207
00:11:15,913 --> 00:11:17,913
但它們回到7分鐘18分鐘的影片

208
00:11:17,913 --> 00:11:19,913
並且提出出一句我提出的

209
00:11:19,913 --> 00:11:21,913
《Specific techniques》會確定改變,

210
00:11:21,913 --> 00:11:24,913
但能力去解釋你的 intent

211
00:11:24,913 --> 00:11:26,913
會不會出現出的風格

212
00:11:26,913 --> 00:11:28,913
它們的這句話中

213
00:11:28,913 --> 00:11:30,913
兩者三者的詞

214
00:11:30,913 --> 00:11:31,913
也能夠適合出的詞

215
00:11:31,913 --> 00:11:33,913
它們也能夠適合出的詞

216
00:11:33,913 --> 00:11:35,913
從開始的第三者

217
00:11:35,913 --> 00:11:38,593
所以第一次的二字幕 remained untouched

218
00:11:38,593 --> 00:11:42,313
整個最終的一段時間是84.9秒長

219
00:11:42,313 --> 00:11:43,793
在它的翻譯文章

220
00:11:43,793 --> 00:11:44,873
它們確實說

221
00:11:44,873 --> 00:11:47,153
它們的詞語詞在前面的詞語詞

222
00:11:47,153 --> 00:11:48,833
不可以保持這個詞語

223
00:11:48,833 --> 00:11:51,793
聽聽說這分比之前和後

224
00:11:51,793 --> 00:11:55,473
重新的心情感

225
00:11:55,473 --> 00:11:58,673
但其實是很簡單的解決

226
00:11:58,673 --> 00:12:02,233
只要是小學生的解決

227
00:12:02,233 --> 00:12:04,673
這根本的解決能不能改變

228
00:12:04,673 --> 00:12:08,033
重新的心情感

229
00:12:08,033 --> 00:12:10,593
學習的解決能確定會改變

230
00:12:10,593 --> 00:12:14,513
但能不能改變解決定的解決能

231
00:12:14,513 --> 00:12:17,233
其實是很簡單的解決

232
00:12:17,233 --> 00:12:20,633
只要是用人的想法,

233
00:12:20,633 --> 00:12:22,833
the core of this ability has never changed.

234
00:12:22,833 --> 00:12:24,833
It didn't just crudely delete sentences.

235
00:12:24,833 --> 00:12:27,033
Instead, it went back to the original footage,

236
00:12:27,033 --> 00:12:28,233
found something I had said,

237
00:12:28,233 --> 00:12:31,133
and completely rounded out the logic of the entire short video.

238
00:12:31,133 --> 00:12:32,633
This really took me by surprise.

239
00:12:32,633 --> 00:12:36,133
This is the kind of truth you'll never see in other people's demonstrations.

240
00:12:36,133 --> 00:12:39,333
It won't discover on its own that your sentences might not connect.

241
00:12:39,333 --> 00:12:41,133
But as long as you point out the flaw,

242
00:12:41,133 --> 00:12:42,933
it can return to the original footage,

243
00:12:42,933 --> 00:12:44,233
find a sentence you said,

244
00:12:44,233 --> 00:12:45,633
and save the short clip for you.

245
00:12:45,633 --> 00:12:49,433
In the past, we had to edit it piece by piece, by hand.

246
00:12:49,433 --> 00:12:52,433
现在,它变成了一个问题的问题,

247
00:12:52,433 --> 00:12:54,053
然后它会去解决它们的问题。

248
00:12:54,053 --> 00:12:56,053
好了,现在,我们的解决了,

249
00:12:56,053 --> 00:12:58,673
我们看看如何练习的手工具能够得到。

250
00:12:58,673 --> 00:13:01,053
当我开始了我的AI editing工作,

251
00:13:01,053 --> 00:13:03,593
最后一切的部分是最痛苦的部分,

252
00:13:03,593 --> 00:13:07,353
在一般的练习,细节的练习,

253
00:13:07,353 --> 00:13:08,813
剪达两句话,

254
00:13:08,813 --> 00:13:10,873
让它们的音乐不停,

255
00:13:10,873 --> 00:13:13,433
操作,或是随着的语言

256
00:13:13,433 --> 00:13:15,013
是非常困难的。

257
00:13:15,013 --> 00:13:18,233
所以,这次,我选择两个字的短片

258
00:13:18,233 --> 00:13:20,233
我和它在一起的字幕中

259
00:13:20,233 --> 00:13:22,233
和一句話中

260
00:13:22,233 --> 00:13:24,233
不斷它從別人的東西

261
00:13:24,233 --> 00:13:26,233
而它開始了

262
00:13:26,233 --> 00:13:28,233
從它們的時間尖

263
00:13:28,233 --> 00:13:30,233
然後,在它們的時尖

264
00:13:30,233 --> 00:13:33,233
它們的小小小小小小

265
00:13:33,233 --> 00:13:34,233
和一句話

266
00:13:34,233 --> 00:13:36,233
而它們的字幕也在一起

267
00:13:36,233 --> 00:13:38,233
之後我聽到它們

268
00:13:38,233 --> 00:13:39,233
我聽到它們的字幕

269
00:13:39,233 --> 00:13:40,233
並不聽到它們

270
00:13:40,233 --> 00:13:42,233
沒有音響和沒有音響

271
00:13:42,233 --> 00:13:43,233
沒有音響

272
00:13:43,233 --> 00:13:45,233
每一個AI的話題

273
00:13:45,233 --> 00:13:48,233
你真的只有三個選擇

274
00:13:48,233 --> 00:13:51,753
每一位 AI人的主題目 出現中的一種

275
00:13:51,753 --> 00:13:53,753
你只有三個選擇

276
00:13:53,753 --> 00:13:56,393
一種主題目的問題 並沒有改變

277
00:13:56,393 --> 00:13:59,673
而是一種技術的問題 並且是一種技術的問題

278
00:13:59,673 --> 00:14:02,753
它是一種主題目的問題 並且是一種主題目的問題

279
00:14:02,753 --> 00:14:06,993
它是一種主題目的問題 並且是一種主題目的問題

280
00:14:06,993 --> 00:14:09,633
它是一種主題目的問題 並且是一種主題目的問題

281
00:14:09,633 --> 00:14:14,033
當然,在 visual上 它們的主題目 也對於是一種主題目的問題

282
00:14:14,033 --> 00:14:21,233
在眼睛不聽,眼睛能看到它,因為我需要它不觸摸到其他相機,

283
00:14:21,233 --> 00:14:24,233
所以它不可以把它測試到另一種景點。

284
00:14:24,233 --> 00:14:27,033
它甚至是在報告中的報告,

285
00:14:27,033 --> 00:14:30,833
而 Cloud Code Ops 5.5也提供了一提提,

286
00:14:30,833 --> 00:14:34,233
如果我還想要加一個4-Frame-Dissolve在剪接,

287
00:14:34,233 --> 00:14:36,233
但我選擇不加這次的一刻。

288
00:14:36,233 --> 00:14:38,633
在短期,它剪掉了一句話中,

289
00:14:38,633 --> 00:14:40,433
在中間中的一種感覺,

290
00:14:40,433 --> 00:14:43,033
這就是一種平常的環境,

291
00:14:43,033 --> 00:14:44,693
and we really can't fault it for that.

292
00:14:44,693 --> 00:14:46,033
After the text editing,

293
00:14:46,033 --> 00:14:48,333
I then had it pick one sentence in the clip

294
00:14:48,333 --> 00:14:50,133
that most needed a visual aid

295
00:14:50,133 --> 00:14:51,533
and add an animation

296
00:14:51,533 --> 00:14:53,933
while leaving the rest of the clip untouched.

297
00:14:53,933 --> 00:14:54,833
It chose,

298
00:14:54,833 --> 00:14:58,733
The scenario may have shifted from simple one word dialogues

299
00:14:58,733 --> 00:15:01,433
to long and complex process chains

300
00:15:01,433 --> 00:15:05,233
because other sentences can be understood through audio and subtitles alone,

301
00:15:05,233 --> 00:15:07,833
but this one explains a structural shift

302
00:15:07,833 --> 00:15:09,933
that truly requires visual support.

303
00:15:09,933 --> 00:15:39,933
所以在36-42秒之中,它用了约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约约�

304
00:16:09,933 --> 00:16:11,933
我认为了一种看法的,我们就会看到了一种看法的,

305
00:16:11,933 --> 00:16:16,173
当然,我们的AI能力导致的竹签是一种独特的,

306
00:16:16,173 --> 00:16:17,933
并且这种程度是一种更多的,

307
00:16:17,933 --> 00:16:21,213
而这次我们将会把整个电影的电影视频放在一种演讲解的,

308
00:16:21,213 --> 00:16:25,213
我们将会把电影视频放在一种演讲解的,

309
00:16:25,213 --> 00:16:30,893
你必须要保持原本的音乐和字幕内并且保持我的拍摄

310
00:16:30,893 --> 00:16:33,453
也保持完整的配置,

311
00:16:33,453 --> 00:16:38,493
它没有任何存在的实际上的实际文字体或 AI image generation tools,

312
00:16:38,493 --> 00:16:42,733
It was hand drawn stroke by stroke with a complete set of storyboards

313
00:16:42,733 --> 00:16:46,573
Dividing the forty subtitle segments into about thirty different shots

314
00:16:46,573 --> 00:16:49,613
Speech bubbles, three cards, a lock and key,

315
00:16:49,613 --> 00:16:52,233
A wrecking ball hitting a house, a wrench to fix it

316
00:17:07,233 --> 00:17:08,233
如果一旦新的AI buzzword

317
00:17:08,233 --> 00:17:09,233
發生在業業

318
00:17:09,233 --> 00:17:11,233
你真的只有三個選擇

319
00:17:11,233 --> 00:17:12,233
第一次選擇

320
00:17:12,233 --> 00:17:13,233
第一次選擇

321
00:17:13,233 --> 00:17:14,233
老問題

322
00:17:14,233 --> 00:17:15,233
老問題

323
00:17:15,233 --> 00:17:16,233
第一次選擇

324
00:17:16,233 --> 00:17:17,233
沒有改變

325
00:17:17,233 --> 00:17:18,233
而解決的

326
00:17:18,233 --> 00:17:19,233
解決定

327
00:17:19,233 --> 00:17:20,233
和解決定

328
00:17:20,233 --> 00:17:21,233
完全一樣

329
00:17:21,233 --> 00:17:22,233
這類型的類型

330
00:17:22,233 --> 00:17:24,233
你會只要用一種AI tools

331
00:17:24,233 --> 00:17:25,233
有必要的一種AI

332
00:17:25,233 --> 00:17:27,233
完全沒有需要

333
00:17:27,233 --> 00:17:28,233
完全遮掩

334
00:17:28,233 --> 00:17:29,233
或是離開

335
00:17:29,233 --> 00:17:31,233
作為新的問題

336
00:17:31,233 --> 00:17:33,233
只要改變了你的習慣

337
00:17:33,233 --> 00:17:34,233
是足夠的

338
00:17:34,233 --> 00:17:36,233
只要是當中的問題

339
00:17:36,233 --> 00:17:38,233
是我們的第三個動作

340
00:17:38,233 --> 00:17:39,233
改變的

341
00:17:39,233 --> 00:17:41,233
但是,有一個小小的問題

342
00:17:41,233 --> 00:17:44,233
在這首歌中提到三個選擇

343
00:17:44,233 --> 00:17:45,233
三個選擇

344
00:17:45,233 --> 00:17:46,233
三個選擇

345
00:17:46,233 --> 00:17:47,233
三個選擇

346
00:17:47,233 --> 00:17:48,233
一秒鐘

347
00:17:48,233 --> 00:17:49,233
它說實在這裏

348
00:17:49,233 --> 00:17:50,233
可以只能夠

349
00:17:50,233 --> 00:17:51,233
去看照相機

350
00:17:51,233 --> 00:17:52,233
而不能夠

351
00:17:52,233 --> 00:17:54,233
其實像一個人

352
00:17:54,233 --> 00:17:55,233
但能夠描述

353
00:17:55,233 --> 00:17:56,233
一整個選擇

354
00:17:56,233 --> 00:17:57,233
一整個選擇

355
00:17:57,233 --> 00:18:01,073
frame by frame, using only code, shows a very high level of completion.

356
00:18:01,073 --> 00:18:05,553
During the entire testing process, not a single task failed or pushed back on me.

357
00:18:05,553 --> 00:18:11,153
My overall assessment of this test is that all tasks were basically free of obvious issues

358
00:18:11,153 --> 00:18:15,553
and completed beautifully. Plus, it only took one prompt each time to finish,

359
00:18:15,553 --> 00:18:20,433
without needing repeated patches or re-rolling. So let's return to the opening question.

360
00:18:20,433 --> 00:18:23,633
Is Claude 3.5 Sonnet really that efficient at video editing?

361
00:18:23,633 --> 00:18:26,593
我的答案是它是非常非常強的

362
00:18:26,593 --> 00:18:28,193
甚至更強的我以為

363
00:18:28,193 --> 00:18:29,633
如果一個作用的作用

364
00:18:29,633 --> 00:18:31,433
能夠穿得到工作的

365
00:18:31,433 --> 00:18:32,833
那會包括選擇

366
00:18:32,833 --> 00:18:34,833
索尼索的長期影片

367
00:18:34,833 --> 00:18:36,433
剪輯切剪輯

368
00:18:36,433 --> 00:18:38,433
剪輯的錄音

369
00:18:38,433 --> 00:18:40,433
控制製作

370
00:18:40,433 --> 00:18:41,633
短片的影片

371
00:18:41,633 --> 00:18:42,433
短片的錄片

372
00:18:42,433 --> 00:18:43,633
短片的影片

373
00:18:43,633 --> 00:18:45,633
甚至更加整理整理

374
00:18:45,633 --> 00:18:47,233
整理的畫面

375
00:18:47,233 --> 00:18:49,433
它遇到所有的重量

376
00:18:49,433 --> 00:18:51,073
在整個效率的

377
00:18:51,073 --> 00:18:52,873
我們是很簡單的

378
00:18:52,873 --> 00:18:56,293
之前開始,明確定你的認識 criteria

379
00:18:56,293 --> 00:18:58,133
當最後的影片是出發,

380
00:18:58,133 --> 00:19:00,633
放在你的耳機上聽聽聽聽聽聽

381
00:19:00,633 --> 00:19:03,753
如果你的音效不流,或是否有效,

382
00:19:03,753 --> 00:19:05,413
就尋找到它們的解決

383
00:19:05,413 --> 00:19:08,753
真實,我會覺得很難以為

384
00:19:08,753 --> 00:19:11,633
當所有的重點是用力的動作是用來的

385
00:19:11,633 --> 00:19:12,413
什麼是留意?

386
00:19:12,413 --> 00:19:14,953
解決什麼是好或是用來的

387
00:19:14,953 --> 00:19:16,913
根本就完全是你的評判

388
00:19:16,913 --> 00:19:19,753
沒有工具能寫這些評論的評論

389
00:19:19,753 --> 00:19:21,753
如果你有些畫面的畫面

390
00:19:21,753 --> 00:19:23,753
如果你想要做一件事的工作

391
00:19:23,753 --> 00:19:25,753
你可以做完整的方法

392
00:19:25,753 --> 00:19:27,753
第一步,不在买长时间的方法

393
00:19:27,753 --> 00:19:29,753
你可以先做一段时间

394
00:19:29,753 --> 00:19:31,753
你可以先做一个短时间

395
00:19:31,753 --> 00:19:33,753
看一段时间,看一段时间

396
00:19:33,753 --> 00:19:35,753
看一段时间,看一段时间

397
00:19:35,753 --> 00:19:37,753
然后让它进行完成

398
00:19:37,753 --> 00:19:39,753
第二步,你可以确定

399
00:19:39,753 --> 00:19:40,753
你接受的规则

400
00:19:40,753 --> 00:19:41,753
之前开始开始做

401
00:19:41,753 --> 00:19:43,753
三个规则

402
00:19:43,753 --> 00:19:45,753
和七个细节 prompts

403
00:19:45,753 --> 00:19:47,753
都在这里

404
00:19:47,753 --> 00:19:49,753
在第三步

405
00:19:49,753 --> 00:19:51,053
如果你的手機是做到,

406
00:19:51,053 --> 00:19:52,373
你必須用手機的聽話

407
00:19:52,373 --> 00:19:53,993
聽到通常的通常

408
00:19:53,993 --> 00:19:54,873
一開始你發現

409
00:19:54,873 --> 00:19:56,993
音樂或音樂不接觸

410
00:19:56,993 --> 00:19:58,413
就請你解釋一下

411
00:19:58,413 --> 00:20:00,133
然後請你去解釋一下

412
00:20:00,133 --> 00:20:02,133
你可以選擇最好的解釋

413
00:20:02,133 --> 00:20:05,053
最後,記得老說

414
00:20:05,053 --> 00:20:07,373
因為AI不表達的錯誤

415
00:20:07,373 --> 00:20:08,513
不代表你會正確

416
00:20:08,513 --> 00:20:10,433
你會永遠負責任

417
00:20:10,433 --> 00:20:11,373
為最後的解釋

418
00:20:11,373 --> 00:20:12,753
更多用力的技巧

419
00:20:12,753 --> 00:20:13,633
更多用力

420
00:20:13,633 --> 00:20:14,133
更多用力的能力

421
00:20:14,133 --> 00:20:15,873
更多用力的解釋

422
00:20:15,873 --> 00:20:17,133
和定義的 standards

423
00:20:17,133 --> 00:20:19,973
我的MAPS course starts exactly from this step

424
00:20:19,973 --> 00:20:21,413
and the link is in the description

425
00:20:21,413 --> 00:20:23,413
alright see you in the next video