貓咪暴揍哥斯拉、葉問拳打鋼鐵俠、孫悟空棒打奧特曼、小學(xué)生隔扣詹姆斯……最近，Seedance 2.0全球爆火，相關(guān)視頻也在不少短視頻平臺刷屏。

Chinese tech company ByteDance has launched the latest version of its artificial intelligence video-generation model, Seedance 2.0, which has quickly gone viral worldwide, stunning creators and prompting some US filmmakers to warn it could upend Hollywood.

upend /?p?end/ 顛覆

近日，字節(jié)跳動最新視頻生成模型Seedance 2.0在即夢、豆包、小云雀等產(chǎn)品開啟內(nèi)測。

它支持圖像、視頻、音頻、文本四種模態(tài)輸入，只需輸入提示詞或上傳參考圖，就能在60秒內(nèi)產(chǎn)出帶完整原生音軌的多鏡頭視頻，大幅降低專業(yè)創(chuàng)作的門檻與成本。

憑借多模態(tài)輸入、超強參考能力等亮點，Seedance 2.0很快引爆網(wǎng)絡(luò)。

不僅國內(nèi)，Seedance 2.0更在海外社交平臺上持續(xù)刷屏、頻頻登上熱搜。海外用戶積極嘗試用它生成諸如“悟空對決漫威英雄、寵物貓狗隔扣NBA球星”等天馬行空的視頻內(nèi)容，多條演示視頻播放量破百萬。

Supporting image, video, audio and text inputs, the model can generate multi-shot videos with synchronized soundtracks in about 60 seconds, significantly lowering production costs and technical barriers. Its advanced reference capabilities and multimodal integration have sparked global interest, with imaginative demo clips on overseas social media platforms drawing millions of views.

美國導(dǎo)演：這可能會搞垮好萊塢

據(jù)參考消息報道，多位美國導(dǎo)演或影視從業(yè)者在使用Seedance 2.0之后，公開表達了諸如“好萊塢要完蛋了”或“這東西說不定真能搞垮好萊塢”的感慨。

Several American directors and film professionals have expressed surprise after testing the tool, with some calling it a potential disruptor of traditional filmmaking.

導(dǎo)演Charles Curran在社交平臺連續(xù)發(fā)文，他為從未面世的真人電影制作了一版預(yù)告片，僅用20分鐘，花費60美元，借助Seedance 2.0完成，“這說不定真能搞砸好萊塢呢”。

在他發(fā)布的1分24秒的視頻里，做到了電影級畫面特效、音畫同步、多機位“拍攝”等。

US documentary filmmaker Charles Curran said he used the model to produce a trailer for an unreleased live-action film in 20 minutes at a cost of $60. The 1-minute-24-second clip featured cinematic visual effects, synchronized audio and simulated multi-camera shooting.

據(jù)了解，Charles Curran是美國紐約資深紀錄片導(dǎo)演、制片人、剪輯師，長期深耕紀實影像與實驗電影創(chuàng)作，同時為國際頂級品牌提供商業(yè)影像服務(wù)。

不少科技工作者和AI研究者也表達了類似的觀點：

另據(jù)南都周刊報道，對于網(wǎng)友提出的可能混淆現(xiàn)實的擔(dān)憂，2月9日，即夢創(chuàng)作者社群中，平臺運營人員發(fā)布消息稱：“Seedance 2.0在內(nèi)測期間收獲了遠超預(yù)期的關(guān)注，感謝大家的使用反饋。為了保障創(chuàng)作環(huán)境的健康可持續(xù)，我們正在針對反饋進行緊急優(yōu)化，目前暫不支持輸入真人圖片或視頻作為主體參考”，并表示平臺深知創(chuàng)意的邊界是尊重，產(chǎn)品調(diào)整后會以更完善的面貌與大家正式見面。

Addressing concerns about AI-generated content blurring the line with reality, the platform said it is making urgent adjustments following beta feedback. It will temporarily disable the use of real-person images or videos as reference material while refining the product.

來源：中國青年報綜合南都周刊、參考消息、藍鯨新聞、21世紀經(jīng)濟報道等

跟著China Daily

精讀英語新聞

“無痛”學(xué)英語，每天20分鐘就夠！