"},"released":{"wt":"2021"},"genre":{"wt":"Text-generation model,text-to-image model"},"license":{"wt":"Proprietary"},"website":{"wt":"{{URL|https://novelai.net/}}"}},"i":0}}]}" id="mwBw">.mw-parser-output .infobox-subbox{padding:0;border:none;margin:-3px;width:auto;min-width:100%;font-size:100%;clear:none;float:none;background-color:transparent}.mw-parser-output .infobox-3cols-child{margin:auto}.mw-parser-output .infobox .navbar{font-size:100%}@media screen{html.skin-theme-clientpref-night .mw-parser-output .infobox-full-data:not(.notheme)>div:not(.notheme)[style]{background:#1f1f23!important;color:#f8f9fa}}@media screen and (prefers-color-scheme:dark){html.skin-theme-clientpref-os .mw-parser-output .infobox-full-data:not(.notheme) div:not(.notheme){background:#1f1f23!important;color:#f8f9fa}}@media(min-width:640px){body.skin--responsive .mw-parser-output .infobox-table{display:table!important}body.skin--responsive .mw-parser-output .infobox-table>caption{display:table-caption!important}body.skin--responsive .mw-parser-output .infobox-table>tbody{display:table-row-group}body.skin--responsive .mw-parser-output .infobox-table tr{display:table-row!important}body.skin--responsive .mw-parser-output .infobox-table th,body.skin--responsive .mw-parser-output .infobox-table td{padding-left:inherit;padding-right:inherit}}
Developer(s) | Anlatan [1] |
---|---|
Initial release | 2021 |
Type | Text-generation model, text-to-image model |
License | Proprietary |
Website | novelai |
NovelAI is an online cloud-based, SaaS model, and a paid subscription service for AI-assisted storywriting [2] [3] [4] and text-to-image synthesis, [5] originally launched in beta on June 15, 2021, [6] with the image generation feature being implemented later on October 3, 2022. [5] [7] NovelAI is owned and operated by Anlatan, which is headquartered in Wilmington, Delaware. [1]
NovelAI uses GPT [ broken anchor ]-based large language models (LLMs) [8] [9] to generate storywriting and prose. [10] It has several models, such as Calliope, Sigurd, Euterpe, Krake, and Genji, with Genji being a Japanese-language model. [11] [12] The service also offers encrypted servers and customizable editors. [10] [13]
For AI art generation, which generates images from text prompts, NovelAI uses a custom version of the source-available Stable Diffusion [2] [14] text-to-image diffusion model called NovelAI Diffusion, which is trained on a Danbooru-based [5] [1] [15] [16] dataset.
NovelAI is also capable of generating a new image based on an existing image. [17] The NovelAI terms of service states that all generated content belongs to the user, regardless if the user is an individual or a corporation. [5] Anlatan states that generated images are not stored locally on their servers. [1]
On April 28, 2021, Anlatan officially launched NovelAI. [6] [18]
On June 15, 2021, Anlatan released their finetuned GPT-Neo-2.7B model from EleutherAI named Calliope, after the Greek Muses. A day later, they released their Opus-exclusive GPT-J-6B finetuned model named Sigurd, [6] [18] after the Norse/Germanic hero.
On March 21, 2023, NVIDIA and CoreWeave announced Anlatan being one of the first CoreWeave customers to deploy NVIDIA's H100 Tensor Core GPUs [19] for new LLM model inferencing and training. [4] [20] [21] [22]
On April 1, 2023, Anlatan added ControlNet features to their text-to-image NovelAI Diffusion model. [23] [24]
On May 16, 2023, Anlatan announced that they named their H100 cluster Shoggy, a reference to H.P. Lovecraft's Shoggoths, which was used to pre-train an undisclosed 8192 token context LLM in-house model. [25]
Following the implementation of image generation, NovelAI became a widely-discussed topic in Japan, with some online commentators noting that its image synthesis features are very adept at producing close impressions of anime characters, [5] including lolicon and shotacon imagery, [26] while others have expressed concern that it is a paid service reliant on a diffusion model, while the original machine learning training data consists of images used without the consent of the original artists. [5] [27] [1] [15] [16] Attorney Kosuke Terauchi notes that, since a revision of the law in 2018, it is no longer illegal in Japan for machine learning models to scrape copyrighted content from the internet to use as training data; [1] meanwhile, in the United States where NovelAI is based, there is no specific legal framework which regulates machine learning, and thus the fair use doctrine of US copyright law applies instead. [1] Danbooru has posted an official statement in regards to NovelAI's use of the site's content for AI training, expressing that Danbooru is not affiliated with NovelAI, and does not endorse nor condone NovelAI's use of artists' artworks for machine learning. [1] [15] [14]
FayerWayer described NovelAI as a service capable of generating hentai. [28] Manga artist Izumi Ū commented that while the manga style art generated by NovelAI is highly accurate, there are still imperfections in the output, although he views these as human-like in a favourable light nonetheless. [29]
In response to the topic of NovelAI, Narugami, founder of the Japanese freelance artist commissioning website Skeb, stated on October 5, 2022 that the use of AI image generation is prohibited on the platform since 2018. [30] Illustrations using NovelAI have been posted on social media and illustration posting sites, and by October 13, 2,111 works tagged with #NovelAI were posted on Pixiv. Pixiv has stated that it is not considering a complete elimination of creations that use AI, [31] though it requires AI-generated posts to be marked as such and allows users to filter them out. [32]
On October 6, 2022, NovelAI experienced a data breach where its software's source code was leaked. [33] [34]
Kenjiro Tsuda is a Japanese actor, voice actor, and film director. His talent agency is ANDSTIR.
Pixiv is a Japanese online community for artists. It was first launched as a beta test on September 10, 2007, by Takahiro Kamitani and Takanori Katagiri. Pixiv Inc. is headquartered in Sendagaya, Shibuya, Tokyo, Japan. As of January 2024, the site consists of over 100 million members, over 115 million illustration and manga and over 21 million novel work posted. Pixiv aims to provide a place for artists to exhibit their illustrations and get feedback via a rating system and user comments. Works are organized in an extensive tag structure which forms the backbone of the website.
Ai Hashimoto is a Japanese actress, fashion model and singer.
Marika Kouno is a Japanese voice actress and singer from Tokyo. She was affiliated with the agency Mausu Promotion before becoming affiliated with Aoni Production. Debuting as a voice actress in 2013, her first main role was in the 2015 anime television series Seiyu's Life! as Rin Kohana. She and the other main cast members of Seiyu's Life! are also members of the music group Earphones. She is known for her roles as Yua Nakajima in Hinako Note, Yumina Urnea Belfast in In Another World with My Smartphone, Mahiro Oyama in Onimai, and Silence Suzuka in Uma Musume Pretty Derby.
Naoki Saito is a Japanese illustrator, manga artist, and YouTuber. He is a regular contributing artist for the Duel Masters Trading Card Game, the Pokémon Trading Card Game, and Hatsune Miku merchandise. He is also the main illustrator and character designer for the game Dragalia Lost. As a manga artist, he wrote and illustrated Baki Domoe from 2010 to 2014. In addition, Saito runs a YouTube channel offering advice on how to improve artwork and published several books based on its content.
Hiroki Kasahara is a Japanese kickboxer, currently competing in the lightweight division of Shootboxing. He is a three-weight Shootboxing champion, having held titles at featherweight (-57.5 kg), super featherweight (-60 kg) and lightweight (-62.5 kg).
The year 2023 is the 30th year in the history of the K-1, a Japanese kickboxing promotion. The year will start with K-1 World GP 2023: K'Festa 6.
Irasutoya is a website operated by illustrator Takashi Mifune that offers gratis clip art illustrations. These works can be used for both commercial and non-commercial applications, but copyright is not waived, there are moral rights-related restrictions on how they should be used, and restrictions to use amount per commercial project.
DreamBooth is a deep learning generation model used to personalize existing text-to-image models by fine-tuning. It was developed by researchers from Google Research and Boston University in 2022. Originally developed using Google's own Imagen text-to-image model, DreamBooth implementations can be applied to other text-to-image models, where it can allow the model to generate more fine-tuned and personalized outputs after training on three to five images of a subject.
Preferred Networks Inc. is a Japanese startup focused on the research and development of deep learning for IoT applications. The company was spun off from Preferred Infrastructure (PFI), founded by Toru Nishikawa, Daisuke Okanohara, and others on March 26, 2014.
Gainax's 1987 debut work Royal Space Force: The Wings of Honnêamise was the first project on which Hiromasa Ogura served as art director; although later noted for creating much of the aesthetic behind the influential 1995 film Ghost in the Shell, Ogura himself in a 2012 interview regarded Royal Space Force as the top work of his career. Working from Yoshiyuki Sadamoto's color scheme and Takashi Watabe's architectural drawings, Ogura then gave a "a sense of life" to the aesthetics of the world setting of Royal Space Force through background paintings created by himself and a staff of 16, including future Studio Ghibli art director Yōji Takeshige, whose first work in the anime industry was on the film. The film's writer and director, Hiroyuki Yamaga, sought to avoid using what he regarded as the usual visual symbolism of anime, and instead wanted Royal Space Force's art direction to express specific times of day and night; Ogura attempted to convey Yamaga's verbal instructions in graphic form.
"Shukusei!! Loli Kami Requiem" is a song by Japanese artist Ui Shigure from her debut album The Rain Doesn't Stop Yet. Featuring vocals by Shigure in her nine-year-old schoolgirl persona and narration by Tamaki Inuyama, it was composed by D.watt and written by Maron, both members of IOSYS. It is musically a denpa song, having lyrics that attack fictional pedophiles which borrow tropes from the lolicon genre.
DLsite (ディーエルサイト), operated by the Japanese company EISYS, Inc. (株式会社エイシス), is an ecommerce storefront website and digital distribution service for downloading and selling a mixture of all-ages and adults-only doujinshi, doujin games, digital manga, light novel e-books, software, computer games, Android apps, and similar goods. As of 2022, DLsite had reached over 9 million global users.
The Dog & the Boy is a 2023 animated science fiction short film directed by Ryōtarō Makihara. Distributed by Netflix and released on YouTube on 31 January 2023, the film follows the friendship between a robot dog and a boy. It became notable for its use of artificial intelligence art to create its background artwork, which was widely negatively received online.
Shōnen no Zankyō is a Japanese manga series written and illustrated by Mitsunori Zaki. The manga was first serialized in Kodansha's shōnen manga magazine Monthly Shōnen Sirius from October 2016 to September 2017. It was then transferred to Kodansha's Palcy manga app, where it ran from May 2018 to September 2021, with its chapters collected in four tankōbon volumes as of July 2019.
"Goodbye Sengen" is a 2020 song written by Japanese music producer Chinozo, utilizing the Vocaloid voice library V Flower. The song is the most viewed Vocaloid song on YouTube with more than 100 million views, and received over 3 billion listens on TikTok by 2022. A light novel adaptation was released on October 25, 2021, and sold over one hundred thousand copies by 2022.
ComfyUI is an open source, node-based program that allows users to generate images from a series of text prompts. It uses free diffusion models such as Stable Diffusion as the base model for its image capabilities combined with other tools such as ControlNet and LCM Low-rank adaptation with each tool being represented by a node in the program.
AUTOMATIC1111 Stable Diffusion Web UI is an open source generative artificial intelligence program that allows users to generate images from a text prompt. It uses Stable Diffusion as the base model for its image capabilities together with a large set of extensions and features to customize its output.
For example, NovelAI has been experimenting with Stable Diffusion to produce art that can accompany the AI-generated stories created by users on its platform.
今天我们将要介绍一款最近比较热门的网站 NovelAI, 它既能自动续写小说, 还能进行绘画.[Today we will be looking at NovelAI, a website making the rounds recently, which can automatically continue writing novels, and also can perform paintings.]
学習したデータセットに海外の二次元画像サイト・Danbooruの画像を多く含んでいるといわれています.そんなDanbooruの公式運営チームが、NovelAIについて声明を発表しました... しかし、Danbooruにアップロードされている二次元イラストの多くがPixivやTwitterなどのSNSに投稿された作品の転載であることが問題となっています... 画像生成部分についてはStable Diffusionにならっていながら、独自のアルゴリズムを実装しているとのこと.[The training dataset contains many images from the overseas 2D image site Danbooru. Danbooru's official management team issued a statement about NovelAI... However, many of the 2D illustrations uploaded to Danbooru are reuploads of works posted on social media such as Pixiv and Twitter... In regards to the site's image generation feature, it implements its own algorithm based on Stable Diffusion.]
海外のイラストサイト「Danbooru」は10月5日までに、画像生成AI「NovelAI Diffusion」(NovelAI)と無関係であると主張する声明を発表した... その中で、NovelAIの公式Twitterアカウント(@novelaiofficial)は、データ学習元にDanbooruを使っていると投稿.Twitter上ではNovelAIの違法性を問う声も上がっており... 声明では「われわれがNovelAIに関係していないことと、彼らがしてることを、支持はおろか容認もしていない」と公表.[On October 5, the overseas illustration website "Danbooru" released a statement stating that it had nothing to do with the image generation AI "NovelAI Diffusion" (NovelAI)... Among them, NovelAI's official Twitter account (@novelaiofficial) posted that it uses Danbooru as its training data. On Twitter, there are also voices questioning whether or not NovelAI is operating illegally... "We are not affiliated with NovelAI, and we do not endorse nor condone what they are doing," the statement said.]
Novel AI 近期在官方推特留言區回覆網友留言時, 提及自家人工智慧有透過名叫「Danbooru」的 ACG 圖片網站進行訓練... 許多圖片多來自 pixiv 跟推特的繪師二創、原創作畫... 隨著爭議引爆, 消息也傳到日本, 引起創作者的關切.[NovelAI recently responded to an online commentator's message on its official Twitter, stating that the AI was trained through the ACG image website "Danbooru"... Many pictures are reposts of fanarts and original artworks from Pixiv and Twitter... As the controversy unfurled, the news also reached Japan, causing concern among content creators.]
Anlatan, developers of NovelAI, will be among the first to deploy the latest NVIDIA H100 Tensor Core GPUs on CoreWeave, which began offering the new instances to select customers in February.
NovelAI, a generative AI company that creates literature and imagery with users' personal perspective and voice, will be one of the first customers to use CoreWeave's NVIDIA HGX H100 instances;
CoreWeave valued at $7 billion
今回は、2023/4/1のNovelAIアップデートで追加された「ControlTools(コントロールツール)」についての記事になります!This time, it will be an article about " ControlTools" added in the NovelAI update on April 1, 2023!
Novel AIが、2023年4月1日に大型アップデートを発表したので、チェックしたいと思います.Novel AI has announced a major update on April 1, 2023, so I would like to check it out.
當然, 引起話題討論最多的是 NovelAI 所產出的大量蘿莉繪圖...這些圖片的題材非常廣泛, 也包含許多令人熟悉的要素, 如少女、天使、泳裝、女王、甚至百合或正太等領域[Of course, widely discussed is the large number of lolicon drawings produced by NovelAI... The subject matter of these images are quite broad, and contain elements such as shōjo, angels, swimsuits, queens, and even lesbians and shotacon]
AIの学習元がイラスト無断転載サイト「danbooru」である点が指摘されています... NovelAIは有料プランが前提となっているため、「金を取るのは流石におかしいのでは」といった声も上がっています.[It has been pointed out that the training material for the AI is Danbooru, an unauthorised illustration reuploading website... Since NovelAI works off a paid subscription, there has also been commentary saying that "it's definitely fucked up to be taking money".]
「Skebでは現在、ツールとしてのAIの利用を一律して認めていない」──イラスト発注サイト「Skeb」を運営するスケブ(東京都港区)の創業者のなるがみ... 画像生成AI「NovelAI Diffusion」(NovelAI)が話題になっていることを受け、AIについて自身の意見を述べている.["Skeb currently rules out permitting the use of AI as a tool", states the founder of illustration commissioning site Skeb (based in Minato Ward, Tokyo), Narugami... In response to the issue relating to image generation AI "NovelAI Diffusion" (NovelAI), he expressed his view on AI.]
そして2022年10月6日、NovelAIの公式Twitterアカウントは、AnlatanのGitHubリポジトリがハッキングに遭い、NovelAIのソースコードが流出したと発表しました.[On October 6, 2022, NovelAI's official Twitter account announced that Anlatan's GitHub repository was breached, and that NovelAI's source code was leaked.]
最も影響が大きかったのは10月8日に起きたとされる「Novel AI」のソースコード流出事件です.どういった形でハッキングされたのかは明らかにされていませんが、流出したとされるコードはおそらく本物だという結論になっています.[The biggest impact was the leak of NovelAI's source code, which reportedly occurred on October 8. It is not clear how the breach occurred, but it is very likely that the leaked code is genuine.]