SORA全AI生成影片席捲視頻生成看完就明白真實到令人懷疑現實【Yolo AI News】

Рет қаралды 60,838

4 ай бұрын

🎙️AI 系列Podcast - bit.ly/YoloAIPodcasts
🎙️Elon 馬斯克系列 Podcast - bit.ly/YoloElonPodcasts
OpenAI SORA 繼ChatGPT與DALL-E之後全AI生成影片再顛覆AI行業標準看完就明白真實到令人懷疑現實顛覆視頻生成
Introducing SORA AI - the revolutionary AI video generator that will change the game for content creators.
SORA - 文字生成影片 - openai.com/sora
不同方式支持Yolo 街 ❤️
🔔去Kurious.Live 訂閱 Yolo 街 kurious.live/en/creator/Yolo%...
❤️‍🔥點擊 Super Chat & Super Stickers
👄喺討論區分享你嘅意見
👍點擊 Like
Twitter - / bonkat
主持：Bonnie

Пікірлер: 120

@KillingJoy 4 ай бұрын

Sora 真係堅! 科技日新月異，人類只可以接受、習慣！想當年阿凡達第一集出時都驚為天人，而家睇3D電影已經冇當時嘅興奮感覺⋯

@tiffanywang9930 4 ай бұрын

Thanks, Bonnie! AI 發展的速度真的令人嘆為觀止。😊 多謝分享。❤

@sailaili8141 4 ай бұрын

可以讓逝者透過AI片重生，安慰在世者心靈

@MrNineChun 4 ай бұрын

有點恐怖

@denniskong1687 4 ай бұрын

國家主席

@loeric1757 4 ай бұрын

咁樣對仍然在生嘅人好唔健康，就算日後做到，人類都唔應該嘗試

@kcchannel3988 4 ай бұрын

聽了兩集講AI. 好有趣。令我download ollama試用下。多謝

@maggieli3535 4 ай бұрын

Thanks, 期待下一集

@lukgot2520 4 ай бұрын

無與倫比的的知識渠道‘應該不只對我而言👍👍👍

@user-wl8cl6nj1e 4 ай бұрын

唔該晒Bonnie

@kamhunglai6920 4 ай бұрын

Thanks Bonnie 讓我知道咁多科技新知識🙏🏾

@yologuy 4 ай бұрын

You're most welcome!

@betttyyip572 4 ай бұрын

Thank you, Bonnie. I am in awe.

@yologuy 4 ай бұрын

Me too!

@27868415 4 ай бұрын

好厲害

@user-zy3jc7nm7h 4 ай бұрын

Nothing is impossible in the future 😎

@Cloudy7102 4 ай бұрын

Thanks ❤

@DoomsdaySpecter 4 ай бұрын

1:48 sora目前並未有處理物理交互的能力 5:40 因為3d動畫中要麼每條毛髮作為一個一個多邊形,要麼用貼圖素材模仿,兩者都沒法做得自然 6:08 其實也有放一些明顯穿崩的片,以展示目前的極限 7:58 無論哪個生成式ai,通常光影都很強 16:23 最易看穿的是文字,特別是漢字 18:06 圖片有inprint功能,再不濟還能ps,但影片的編輯成本則高得多(只能逐幀手動修) 20:36 先不說sora會在prompt層面攔截使用名人,其實真要做謠言,目前來說肯定是真人拍攝+deepfake更好

@yologuy 4 ай бұрын

該模型很可能不使用任何確定性物理規則。更可能的是透過從海量資料集中學習來內化物理學。

@DoomsdaySpecter 4 ай бұрын

@@yologuy 這跟遊戲3D製作的歷史進程有幾分相似,最初的遊戲3D都是不帶物理引擎,所有運動全是人手控制

@jacp1021 4 ай бұрын

叫他重拍小美人魚就好了。😊

@mafai-zf3ri 4 ай бұрын

已經可以呃到大部份人，特別係老人家

@niccoloo0 4 ай бұрын

講得幾好！加油！支持廣東話創作。

@yologuy 4 ай бұрын

Weeee!

@curtlam4033 4 ай бұрын

Thank you Bonnie. ❤ 恭祝龍馬精神身體健康笑口常開繼續拍多啲好片畀我哋呀

@yologuy 4 ай бұрын

Thanks Curt! Wishing you health and happiness in the year of the Dragon.

@mancheonghoho 4 ай бұрын

過去數十年科技和現在AI科技進步仍集中於提升視覺、聽覺效果，如果能虛擬真實觸覺，味覺，那才是大突破！試想那時沒有真的吃東西，味覺上卻像吃了一餐美食，沒有和真人做愛，觸覺上像真的做了，不需要實物而能滿足官感的人生多幸福！

@MaR-ni4qo 4 ай бұрын

近年科技進步嘅速度，真係快到有啲驚

@SkLui-rp1dd 4 ай бұрын

吾己老矣，下世代真係會被科技取代 ! 😎🤔

@yologuy 4 ай бұрын

Never give up! =) Our brains are incredible if you don't let it think it's about to retire!

@Colala2046 4 ай бұрын

Amazing!, 科技發展的日新月异、令人惊叹不已！谢谢你带给我们的知識！在学習中進步！

@keichow9271 4 ай бұрын

😮 演藝學院、藝人的末日倒數……. 嗎？可能受環保人仕歡迎♻️因為不需要用完掉棄的真道具！

@daydaylaugh796 4 ай бұрын

nice

@TTC1980 3 ай бұрын

版主講解得好出色

@yologuy 3 ай бұрын

Thank you 四哥😇

@ivylam3540 4 ай бұрын

打擊廣告業

@ai8989889 4 ай бұрын

唔知幾時會打擊AV行業

@carsonk3366 4 ай бұрын

❤

@simonwong3831 4 ай бұрын

我以使用 Midjourney 生成圖像的經驗，相信現在 OpeAI SORA Demo, 絕不是一段小文字，一次便能生成的。是經過無數次修正改良，才有現在的表現。其實 Time cost都是非常高！

@iamrobbechoi 4 ай бұрын

Bonnie, 你的見解非常特別,keep up with the great work! 我是不知道這邊平常有多少人在接觸AI的技術，我算是有。 Sora有多可怕，就是大家都知道AI能生成影片，也在努力讓AI生成影片，但大家都是幼稚園小學生程度左右的時候，突然冒出了一個大學畢業生... 最可怕的不是Sora算出來的影片有多強，而是生成式AI才多久的時間已經能達到這樣的程度，甚至動畫生成也就是去年才開始有人做的，今天就出來一個完成度到這個程度的東西。我早前用AI生成些影片,有空請看看及指教🙏 kzbin.info/www/bejne/qKCweKevpsSUotksi=gWxfbD_E8MC8o7C4

@yologuy 4 ай бұрын

Thank you for sharing your view point. Some say seeing is believing, I say doing is believing =) Love your ads!

@user-wl8cl6nj1e 4 ай бұрын

AI 相同片so far 仲有一個問題，就係裡面嘅文字都係一堆無意思嘅類文字符號，不過，呢個問題相信會改善。

@user-tz9uc8gh4f 4 ай бұрын

Sora 可以拍天堂地獄….

@ckckck633 4 ай бұрын

好多行業夕陽西下

@joe3223 4 ай бұрын

好快可以做動畫電影

@mansingcheng7020 4 ай бұрын

曾有預言指人類最終被 AI 打敗、甚至滅絕，那並非不可能! 只能祈求有生之年不會發生。

@tikyu2801 4 ай бұрын

Terminator 好快到

@Papaman-no5rh 4 ай бұрын

其實要入d真片令AI進化😮

@kauffmann101 4 ай бұрын

AI 鍊金術師

@abrahamyuen8621 4 ай бұрын

嘆為觀止之餘，亦不無驚恐；不在今天，但其實未遠。

@donaldchin3140 4 ай бұрын

好嘅開始，但未成氣候。

@bentsang3456 4 ай бұрын

俾小小建议，讲解其实好好，不过希望可以流畅小小，听起身会舒服好多~加油！

@yologuy 4 ай бұрын

my Chinese -> 仍在努力中 =)

@getreloaddds8028 4 ай бұрын

無我相，無人相，無眾生相，俾紅藍葯丸我揀吖

@kenxchoi775 4 ай бұрын

kakui ai終於可以有mv🥲

@mychan2600 4 ай бұрын

真係犀鳩厲

@ivylam3540 4 ай бұрын

Hi ❤

@yologuy 4 ай бұрын

Hello 😊

@tansiyitan 4 ай бұрын

人人都是導演

@siriusinkheart 4 ай бұрын

Creating a full-feature movie might still take some time, given the complexities involved. However, I believe a full-length novel generated entirely by AI, with minimal human editing, will be achievable within 10 years, largely due to current limitations in chip bandwidth and computational power. On the other hand, a blockbuster novel produced through a human/AI collaboration could be a reality within the next 5 years. Once we have a full-length novel, another AI program could transform the novel into prompts for a program similar to Sora, which would then stitch together short films into a cohesive full-length movie. These advancements might occur simultaneously with the progress mentioned earlier. Consequently, within 10 years, AI could potentially independently generate a full-length movie worth watching. A few years ago, I stumbled upon a revelation: I firmly believe that everything can be distilled into a domain language. Observing the evolution of LLM, I am increasingly convinced of the validity of this assumption. Perhaps, in a surprising twist, the human brain itself operates akin to an LLM.

@yologuy 4 ай бұрын

Thank you for sharing this! Gives me some food for thought. The more we see these AI developments, the more I am convinced if we can use it to also unlock our understanding of our own brain, how our neural networks function at the most detail level. Then, we will no longer have the saying that we humans are only using 10% of our brain :)

@oliverlui9181 4 ай бұрын

謝謝，Bonnie。原則上，是否可以用佢將整部金庸小說拍成電影？

@yologuy 4 ай бұрын

In principal, yes. Check out Google's Genesis 1.5 announcement. It can easily take the entire novel as input, help create prompts, which you can then feed into Sora 😎

@mancheonghoho 4 ай бұрын

當AI製作成本比真人低，最先失業的應該是明星、演員、佈景、服裝等工作人員吧！

@everythingispiggy169 4 ай бұрын

太空人條片真係好誇張😂，但目前黎講要取代資源庫未必可能，一黎要生成咁細緻影片係勁吃硬件資源，有財力公司或政府機構先有可能負擔得起，一般人除左不計工本，仲要花好多時間等佢 rende圖；另外係佢要做到越真，需要嘅資源庫數據就越多，除非佢已經建立左1:1地球實境模型😂

@yologuy 4 ай бұрын

You'll be surprised - that is exactly one of the things Nvidia is doing - making a model of the entire earth =)

@edmondhung1231 4 ай бұрын

Sora生產3d影像配合vision pro使用會變成點。。。

@yologuy 4 ай бұрын

Can solve Vision Pro's issue of not having enough content!

@fenix20075 4 ай бұрын

覺得可以用AI生成大堆指令來控制unreal engine 5 生成影片，應該會更加完美……

@yologuy 4 ай бұрын

在Sora之前, 最大嘅挑戰係傳統嘅影片渲染方法。你造3D物件，用渲染引擎處理，並定義相機移動來生成2D影片。Epic Games嘅Unreal Engine 5係呢方面嘅一個受歡迎工具。雖然有效，但依然需要放置同渲染3D物件，同時模擬光線physics。呢種「由3D生成2D」嘅方法一直穩步改進。但由於根本上受限於3D空間同物件嘅概念，所以仍然有限制。而OpenAI嘅「Sora」或許就係一個遊戲規則改變者。佢唔用傳統方法，而係直接從大量嘅影片數據中學習。想像佢吸收咗多年嘅電影同動畫，不知點解內化咗物理嘅規則。但由於 SORA 不是開源的，在 OPenAI 之外，我們還不知道它的模型是如何訓練的。希望隨著時間的推移我們將獲得更多資訊。

@fenix20075 4 ай бұрын

@@yologuy 相反就係佢應該到極限，永遠無法解決gen錯圖既問題，事關係佢用已有既圖片進行學習，於是六隻手指等stable diffusion問題又會出現。然而，如果unreal engine 可以透過AP控制，用程式==文字去生成物件同擺放鏡頭，反向就會變成用prompt生成程式碼來控制unreal engine，錯既機率就會好細，而且可以靠unreal engine執返正。但當然，係要有人可以將大量已有既unreal engine 製作電影既工序變返做文字描述先得。不過認真講，我睇完都唔係幾明，點解Sora可以做到反射，事關學習現有圖像應該好難生成準確既反射圖，因為被反射既物件並無被 3D 化，而圖像學習唔可能學習到其睇唔到既一面。

@yologuy Ай бұрын

我同意你嘅睇法，認為用AI去輔助控制Unreal Engine嘅輸入係可行嘅方案。通過控制Unreal Engine，用文字生成程式碼嚟控制物件同擺放鏡頭，可以大大減少錯誤嘅機率。Right now, I do see a lot of people trying to use AI to replace all kinds of workflows at different companies. I'm sure there're someone working on game production :)

@NgaiChun 4 ай бұрын

可以製作draft trailer, 但一旦製作成品，就像dall-e 3一樣要求製作準確圖片/影片就欠奉了，似乎唔協調/所謂瑕疵，更似係防偽/防deepfake嘅tag….

@colala4481 4 ай бұрын

就算可以做到100%完全真實完美無缺，我都唔覺得人人都可以做到導演或製片，只會提高專業門檻。或者換個講法，真正最受用嘅只會係業界本身，而唔係for平民踏入專門領域。

@mafai-zf3ri 4 ай бұрын

錯重點

@Colala2046 4 ай бұрын

反反复复看了好几遍，太不可思议了😅

@tfc7132 4 ай бұрын

製作沒影響，影響只有演員。不能用明星無所謂，AI能創更完美並且不會衰老的明星😂😂😂

@KamFung 4 ай бұрын

都唔係架, 攝影/燈光/化粧....全部無工開.

@daydayholiday6453 4 ай бұрын

好奇一問，哩條片有冇可能都係全AI生成…？

@yologuy 4 ай бұрын

All the clips in this video are created by Sora =)

@kenxchoi775 4 ай бұрын

但多咗好多假嘅“歷史”片出嚟，到唔知幾年之後，真嘅無人記得，即係又多咗批“Mandela effect”嘅記憶出現

@iloveeat3000 4 ай бұрын

會唔會遲D 藝人演員都冇得撈?

@unknownfolder9329 4 ай бұрын

視頻😑 666 呀, 操真利害, YYDS

@kauffmann101 4 ай бұрын

AI generate 出黎班人o既手痙攣咁😹

@choppar23 4 ай бұрын

2025年之後，我哋重發唔發現到Sora嘅片嘅真假呢？

@yologuy 4 ай бұрын

We will train another AI to help us do the verification :)

@rmnrfbw 4 ай бұрын

將來回到文科世界?

@SKleung473 4 ай бұрын

以後唔方有人再去睇電影, 費事比人呃

@ivylam3540 4 ай бұрын

廣告業

@alberttsang1584 4 ай бұрын

似夢境。似鬼片

@nickwong1321 4 ай бұрын

廣告片為例部分外國景AI做部分人物產品鏡頭自己拍都慳唔少

@Handel-Lok 4 ай бұрын

係咪連人都係ai生成？

@yologuy 4 ай бұрын

看看Apple Vision Pro的虛擬形象功能 =) 老實說 bit creepy

@davedeloera318 4 ай бұрын

AI不會畫手的問題看來還沒解決

@yologuy 4 ай бұрын

Not perfect yet :)

@rylum4921 4 ай бұрын

🦾

@shunchan4411 4 ай бұрын

大家預備好退休生活未？？

@yologuy 4 ай бұрын

Someday we'll all look like the humans in Wall-E 😂

@canwinli9318 4 ай бұрын

所以我们这个世界也是虚拟的吗？

@yologuy 4 ай бұрын

大家開始懷疑人生 =)

@littlebritain 4 ай бұрын

當然唔否認""將來"" (唔知幾時) , 講句說話會有條片出嚟 , 但係啲死乞兒客戶, 睇完你條片就以為撳個掣就有... 又會好多人柒撚懵懵，以為依家撳個掣就有條片(我睇KZbin有人話嘅...).. 未普及到連你都做到嘅時候, 唔好講到咁大... 好多年前宣傳5G 時, 都話,幾秒鐘下載套電影啦, 你睇依家, 我睇唔到5G有幾咁改變世界....

@yologuy 4 ай бұрын

"死乞兒客戶"..... 😂.... i've heard that from my friends in the agency industry.... very frustrating indeed.