imported>Kiyoteru Awaji：/* 外部連結 */

2025-11-03T18:03:24Z

外部連結

新页面

{{Infobox software
| name = Imagen
| logo =
| logo caption =
| screenshot = Scenic_Valley_in_the_Afternoon_%28Imagen%29.jpg
| caption = 由 Imagen 3 所生成的圖像。部分提示語：<code>柔和光照的午後山谷與蜿蜒的河流。</code>
| author =
| developer = [[Google DeepMind]]
| released =
| programming language =
| operating system =
| genre = [[文字轉圖像模型]]
| license =
| repo =
| website = {{URL|https://deepmind.google/technologies/imagen-3/}}
| screenshot size = 250px
| latest release version = Imagen 3
| latest release date = {{start date and age|df=y|2024|8|13}}
}}

'''Imagen'''、'''Imagen 2'''和'''Imagen 3'''是由[[Google DeepMind]]開發的[[文字轉圖像模型]]。在2023年4月[[Google Brain]]與[[Google DeepMind]]合併前，該模型由Google Brain 負責開發。<ref>{{Cite web |title=Google's big AI push will combine Brain and DeepMind into one team |url=https://www.theverge.com/2023/4/20/23691468/google-ai-deepmind-brain-merger |website=[[The Verge]] |date=April 20, 2023 |last1=Roth |first1=Emma |last2=Peters |first2=Jay |url-status=live |archive-url=https://web.archive.org/web/20230420234052/https://www.theverge.com/2023/4/20/23691468/google-ai-deepmind-brain-merger |archive-date=April 20, 2023 |access-date=March 18, 2025}}</ref>Imagen主要用於從文字提示生成圖像，類似於[[Stability AI]]的[[Stable Diffusion]]、[[OpenAI]]的[[DALL-E]]或[[Midjourney]]。

該模型的初代版本最早於2022年5月的一篇論文中介紹。<ref name=":0">{{cite arXiv |eprint=2205.11487 |class=cs.CV |first1=Chitwan |last1=Saharia |first2=William |last2=Chan |title=Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding |date=2022 |last3=Saxena |first3=Saurabh |last4=Li |first4=Lala |last5=Whang |first5=Jay |last6=Denton |first6=Emily |author7=Seyed Kamyar Seyed Ghasemipour |author8=Burcu Karagol Ayan |last9=Sara Mahdavi |first9=S. |author10=Rapha Gontijo Lopes |last11=Salimans |first11=Tim |last12=Ho |first12=Jonathan |author13=David J Fleet |last14=Norouzi |first14=Mohammad}}</ref>它能產生高品質圖像，目前任何擁有Google帳號的使用者皆可透過[[Gemini]]、ImageFX和Vertex AI等服務使用此工具<ref name=":2">{{Cite web |title=Anyone With a Google Account Can Try Google's Latest AI Image Generator Right Now |url=https://lifehacker.com/tech/you-can-try-googles-latest-ai-image-generator-right-now |website=Lifehacker |date=2024-08-16 |language=en |last=Peterson |first=Jake |access-date=2025-03-18}}</ref>

== 歷史 ==
Imagen的初代版本於2022年5月首次在一篇論文中發表，具備從自然語言生成高保真圖像的能力。<ref name=":0" />第二代Imagen 2發佈於2023年12月，<ref>{{Cite web |title=Imagen 2 - our most advanced text-to-image technology |url=https://deepmind.google/technologies/imagen-2/ |website=Google DeepMind |date=2025-03-12 |language=en |access-date=2025-03-18}}</ref>其突出特點是能生成文字與標誌圖像。<ref>{{Cite web |title=Google debuts Imagen 2 with text and logo generation |url=https://techcrunch.com/2023/12/13/google-debuts-imagen-2-with-text-and-logo-generation/ |website=TechCrunch |date=2023-12-13 |language=en-US |last=Wiggers |first=Kyle |access-date=2025-03-18}}</ref> Imagen 3則於2024年8月推出，<ref>{{Cite news |title=Google opens access to Imagen 3, its latest model for AI image generation |url=https://9to5google.com/2024/08/16/google-imagen-3-launch/ |last=Schoon |first=Ben |work=9to5Google |date=2024-08-16 |language=en-US |access-date=2025-03-18 |archive-url=http://web.archive.org/web/20240818012446/https://9to5google.com/2024/08/16/google-imagen-3-launch/ |archive-date=2024-08-18}}</ref>Google表示此新版在細節與光影呈現方面有明顯提升。<ref name=":1">{{Cite web |title=Some of the most realistic AI images you'll see were created with this free tool |url=https://www.techradar.com/computing/artificial-intelligence/what-is-imagen-3-everything-you-need-to-know-about-googles-text-to-image-model |website=TechRadar |date=2025-02-26 |language=en |author1=Christian Rowlands |access-date=2025-03-18}}</ref>

== 技術 ==
Imagen 運用了兩項關鍵技術：

第一，採用了基於[[Transformer架构|Transformer架構]]的[[大型語言模型]]，尤其是{{tsl|en|T5 (language model)|T5}}，用以理解文字並對其進行編碼，供圖像生成之用；

第二，使用階層式擴散模型進行高保真圖像生成。其生成流程分為三個階段：先產生64x64的基礎圖像，接著依序升級至256x256與1024x1024。<ref name=":0" />

== 功能 ==
Imagen可依文字提示生成寫實風格圖像。<ref name=":2" />它也支援多種風格，包括電影感、35毫米膠片風、插畫風和[[超現實]]風。該模型可輸出五種畫面比例：9:16、3:4、1:1、4:3、16:9。此外，Imagen還可透過修改文字提示來編輯已生成的圖像。<ref name=":1" />

== 參見 ==

* [[人工智慧藝術]]
* [[生成艺术|生成藝術]]
* [[Midjourney]]
* [[DALL-E]]
* [[穩定擴散]]

== 參考資料 ==
<references />

== 外部連結 ==

* [https://deepmind.google/technologies/imagen-3/ Imagen 3 website]
* [https://deepmind.google/technologies/imagen-2/ Imagen 2 website]

{{Google AI}}

[[Category:人工智慧藝術]] [[Category:2022年軟體]]

Imagen - 版本历史

imported>Kiyoteru Awaji：​/* 外部連結 */

imported>Kiyoteru Awaji：/* 外部連結 */