Hematemesis sorts out 100 AIGC artifacts, and the workers speed up! Collection is strongly recommended!

Source: Wisdom

Author | Wu Feining

Editor | Li Shuiqing

Original title: "The King of Rolls is Using It! 100 treasure-level AIGC tools to share, highly recommended for collection! ! "

Speaking of AIGC tools, do you still only know ChatGPT?

In fact, more and more AIGC applications are not based on OpenAI's GPT. In the field of entertainment, the cover music of singers "AI Stefanie Sun" and "AI Jay Chou" became popular, and the singers themselves were amazed; in the field of e-commerce, AI virtual humans read scripts generated by AI 7x24 hours, and sold millions of goods for enterprises ; In special classrooms, virtual teachers output sign language courses based on voice content to help deaf people learn knowledge... These scenarios have become the landing field of large-scale self-developed models or open source models.

According to the statistics of Zhishi, as of mid-May, there are at least 100 representative AIGC tools in the world. From daily office work to social media, from game production to graphic design, from financial regulations to product sales, the role positioning of AIGC tools has been upgraded from the previous "observation + prediction" to today's "generation + decision-making", promoting the implementation of AIGC "odd" point" appears.

▲The word cloud map shows around ChatGPT, other AIGC tools with high usage rate and mention rate

In the field of text writing, AIGC tools led by conversational chatbots such as ChatGPT and Wenxinyiyan save people’s time and cost in retrieving information, and can complete a series of inefficient and repetitive tasks in the form of dialogue. Other writing tools also Including Notion AI, Tencent Wenyong, WPS Smart Writing, etc.

In the field of image generation, AIGC has also subverted and reshaped the previous drawing method. "Yiwen Shengtu" provides creators with different styles and endless sources of inspiration, and has set off a revolution in productivity in the field of design. Tools such as Midjourney and DALL-E continue to expand the boundaries of people's imagination.

In the field of audio and video, AI can generate audio and video according to preset styles by analyzing massive source data, which not only shortens the creation cycle, but also breaks through the limitations of physical space and time. Commonly used tools include Xunfei Hearing, MusicLM, Runway Gen-2, etc.

In addition to the above application scenarios, there are also more subdivided scenarios such as collaborative office, language learning, e-commerce live broadcast, programming, and digital human virtual idols. AIGC technology can be used from the technical level with low marginal cost and high efficiency. way to meet the individual needs of users.

What's more worth mentioning is that in the current era of phishing information and fake news flooding the pages, in order to distinguish AIGC content from real content, NetEase and People's Daily Online have successively launched AIGC content detection tools to control content risks. There are also tools related to AIGC content detection abroad, such as Copyleaks, which specializes in text plagiarism detection, and DetectGPT, a plagiarism check assistant for papers.

**This article interprets more than 100 AIGC tools from the seven sections, and attaches web links to help users improve work productivity. **Actually, AIGC tools at home and abroad are springing up like mushrooms every day, so the 100 AIGC tools included in this article do not fully cover the industry, but we hope to provide some reference for the industry in terms of categories and directions.

01. AIGC writing tools: one-click writing

Suitable for life or office scenes

Text generation is one of AIGC's first commercial technologies, and it is also the most mature technology AIGC has developed so far. Today, AI writing tools have made a qualitative leap in the ability to understand context, capture commonsense knowledge, generate long texts, and complete, accurate, and logical content. .

The main landing scenarios of AI writing tools can be roughly divided into the following three categories:

The first is application-oriented text generation, such as sentence search according to meaning, reverse dictionary, etc., which have relatively clear function usage scenarios, and the use direction is also relatively clear. The second is creative text generation, such as Notion AI, WPS intelligent writing, etc., which can be used for text continuation and content generation, etc. Most of them are unstructured writing, and users have greater space and freedom for text creation. The third is conversational text generation, such as Wenxinyiyan, Tongyiqianwen, Xunfei Xinghuo, etc., which are highly interactive and have higher requirements for natural language understanding capabilities of large models.

1. Notion AI: Use ChatGPT to help text "beauty"

Notion AI is a writing assistant whose main functions include writing, editing, summarizing, etc. It can automatically generate blog posts, meeting schedules, social media copywriting, press releases, sales emails, and poetry to meet the needs of different scenarios. Users can let Notion AI process the first draft of an article to get more writing ideas; or use it as an editor to check spelling, grammar, and translation errors.

The tool currently adopts the "free trial + payment" model, providing each new user with 20 free trial opportunities, and after the number of times is used up, you need to purchase the service. The price is 10 US dollars/person/month, which is equivalent to 68.9 yuan.

Web links:

2. Baidu Wenxin Yiyan: "AI joker" who knows Chinese best

Wenxinyiyan is a chat robot developed by Baidu. Its main functions include dialogue and interaction with users, answering questions, and assisting in creation, etc., to help users obtain information, knowledge, and creative inspiration. The usage scenarios include literary creation, business copywriting, mathematics and science. Calculation, Chinese interpretation, multi-modal generation, etc.

In addition to copywriting, the advantages of Wenxinyiyan include the ability to create pictures, and the ability to automatically generate videos based on copywriting.

Web links:

3. Ali Tongyi Qianwen: Writing love letters is easy

Tongyi Qianwen is a large-scale self-developed model of Alibaba Cloud. It is currently equipped with 9 applications, which are mainly divided into efficiency, life and entertainment.

(1) Efficiency category, including three applications: outline writing, SWOT analysis, and product description generation;

(2) Life category, including three applications: "Flying Recipes", "Elementary School Composition", and "And Then";

(3) Entertainment category, including "Rainbow Fart Expert", "Write Love Letters" and "Write Poems for You".

At present, the main functions of Tongyi Qianwen include copywriting, dialogue and chat, knowledge question and answer, logical reasoning, code writing, text summarization, and image and video understanding services.

Web links:

4. Xunfei Xinghuo: 7 major dimensions of ability, the performance is not inferior to ChatGPT

Xunfei Spark is a large model launched by iFLYTEK on May 6. It has seven dimensions of text generation, language understanding, knowledge question and answer, logical reasoning, mathematics, code, and multi-modality. After evaluation and comparison, it is found that its It has outperformed ChatGPT in language comprehension and mathematics ability.

Xunfei Xinghuo can complete multi-style, multi-language, multi-task long text generation, and can also perform grammar detection and error correction on English copywriting, and its language comprehension ability is not inferior to existing systems that are measurable in China.

Web links:

5. Sequence monkey: an AI monkey that can answer complex questions

The large language model "Sequence Monkey" launched by the AI company Mobvoi, its capability system takes language as the core and covers six dimensions of "knowledge, dialogue, mathematics, logic, reasoning, and planning". It can simultaneously support text generation, image generation, Different tasks such as 3D content generation, speech generation and speech recognition.

Sequence Monkey already has a certain ability of natural language understanding, knowledge, logic, and reasoning. For "Which provincial capital has the largest population, Hunan or Hubei?" "Which school did the founder of the company behind Tmall graduate from?" etc. It has been able to quickly give accurate results for such questions that require further thinking.

Web links:

openapi.mobvoi.com

6. Tencent Wenyong Effidit: an artifact of paper writing for wireless continuation

Wenyong Effidit (Efficient and Intelligent Editing) is an intelligent writing assistant developed by Tencent AI Lab. It uses AI technology to assist writers to diverge ideas, enrich expressions, and improve the efficiency of text editing and writing. Its functions include intelligent correction Error, text completion, text rewriting, text expansion, word recommendation, sentence recommendation, generation and other functions.

Web links:

7. Look up sentences according to the meaning of WantQuotes: Encyclopedia of famous quotes

It is a copywriting processing tool developed by the research team of Tsinghua University. It uses the most cutting-edge AI and natural language processing (NLP) technology to help people process reading, writing, copywriting search, and famous quotes more conveniently and quickly. .

Users only need to input relevant subject vocabulary, and it can find relevant famous sayings, poems, sayings, idioms, etc.

Web links:

8. Reverse dictionary WantWord: a dictionary of synonyms and synonyms, farewell words are not expressive

The reverse dictionary and sentence search by meaning are both developed by the research team of Tsinghua University. They can help find more appropriate and vivid synonyms through the given words, and also support simultaneous and mutual translation between Chinese and English.

Web links:

9. FlowUS AI: network disk + memo + writing assistant

Xiliu is a knowledge management and collaborative office software that focuses on providing services for small organizations and individuals. It integrates multiple functions such as writing documents, knowledge storage, multi-dimensional tables, and mind maps into one platform. Its users are college students The group is the main group, accounting for more than 1/3 of the total number of users.

FlowUS has also been adapted to ChatGPT. Users can use FlowUS AI to realize writing, continuing writing, translation, polishing and other functions according to their own needs.

Web links:

10. WPS intelligent writing: automatically generate various articles in 1 second

WPS Smart Writing is an intelligent writing product launched by Kingsoft Office to help users create efficiently. It mainly includes four functions: automatic text generation, auxiliary draft writing, intelligent sentence supplementation, and intelligent proofreading of text.

Its text data and related information come from authoritative media and government public websites. The subject matter covers various writing scenarios such as speeches, summaries, plans, news, etc., and it is connected with Jinshan documents, which can realize the simultaneous uploading of texts to the cloud. After online writing, users can go to Kingsoft Documents performs more professional document editing such as typesetting.

Web links:

11. GrammarlyGo: online grammar "bug catcher"

The English spelling check tool Grammarly also launched the AI service GrammarlyGo, which can generate email drafts based on keyword prompts entered by users, or help existing articles change the tone and text style, adjust the length of articles, etc., and draft outlines for topic writing .

Web links:

12. Volcano Writing WritingGo: One-click translation and polishing

Volcano Writing is a writing assistant launched by ByteDance. It currently supports AI smart writing services for full-text editing. Whether it is revising papers, polishing resumes, writing application documents for studying abroad, writing self-media copywriting, etc. More than 20 writing scenarios, Volcano Writing can be covered.

The user enters the text content that he wants to polish and modify, and after clicking "one-click optimization", it can automatically identify the text type, style and writing purpose. The user can also adjust the extent of modification, and the platform can output it based on the original text with one click. The English rewriting result also supports AI functions such as intelligent error correction and various rewriting, making the language expression more authentic and concise.

Web links:

13. Zhishi Q&A: intelligent Q&A robot

Zhishi Q&A is an intelligent Q&A system based on AI technology. Users can input questions on the Zhishi Q&A platform, and the system will automatically analyze the questions and give the best answer. At the same time, it also provides a variety of interactive methods, including text input, voice input, etc., to meet the needs of different users.

Web links:

In addition to the above familiar AIGC writing tools, there are still many "unpopular products" waiting for user experience, such as Friday AI Writing Assistant, Love Rewriting, Claude, Creator, Secret Tower Writing Cat, Subtxt, Writesonic and so on.

02. AIGC image tool: Vincent's map is more than Midjourney

Freedom to paint with one click

2022 can be said to be the "first year of AI painting". A variety of AI painting tools have demonstrated good image understanding and generation capabilities with the help of text prompts.

With the help of GPT-4, a new wave of competition has also been set off in the field of "Vincent map". Midjourney, which has been updated to the V5 version, is popular all over the Internet with a group of couple photos. Adobe, the leader in the design industry, is not far behind. It hastened to launch "Adobe Firefly" to compete. The majority of design workers. Today's AI image tools are more mature and more varied in terms of commercialization and artistry than last year.

▲ A group of retro couple photos automatically generated by Midjourney

The technical scenarios of image tools can be divided into three types: image generation, image partial modification, and image editing.

One is image generation. Products represented by Midjourney, Stable Diffusion, and DALL-E 2 mainly focus on end-to-end image generation, which can generate a complete image with a specified style based on a text description or a sketch. The underlying technical logic is clear and can provide Creators provide certain sources of inspiration and creative references.

The second is image editing. The main functions include intelligent image watermark removal, setting style filters, modifying image style or improving image clarity, etc., represented by products such as Imagen AI and Chuangketie AI Painter.

The third is partial image modification, the representative product is Adobe Firefly. Its main advantage focuses on changing some elements of the image, or modifying and adjusting layer by layer, which is suitable for secondary creation or post-production improvement.

1. Disco Diffusion: Draw with your mouth

Disco Diffusion is a drawing program that runs on Google Colab. Users with a Google account can run it directly on the browser, but users need to have certain code knowledge.

After the user enters a description sentence, the program can automatically render and generate a picture of the corresponding scene. It is better at generating abstract pictures with a more dreamy style, and the effect is average when generating realistic representations and inputting more text descriptions.

Web links:

2. Midjourney: AI "photographer"

Midjourney is an AI painting chatbot launched by one of the authors of Disco Diffusion, which is carried on Discord. After the launch of GPT-4, it also quickly changed to the V5 version. The new version is more refined in terms of image fidelity and detail processing, and has a higher level of commercialization, almost reaching the point where it can "disguise the real".

In the previous version, the style of generated pictures was mostly cartoon or surreal, and there were few realistic pictures. After updating to the V5 version, Midjourney quickly became popular with a group of photos of couples that were hard to distinguish between true and false, and has reached the The texture of the movie is more realistic in terms of hand close-ups, eye close-ups, and light and shadow processing.

In addition, in Midjourney V5, users can customize the aspect ratio. When entering description text, more detailed adjectives and image details such as mood, style, and light and shade are required. This requires users to have more active control over images and clearer imagination.

Web links:

3. Stable Diffusion: pixel-level image generation

Stable Diffusion is a free and open-source AI image generator. Currently, the latest version of Stable Diffusion XL has been tested for the public.

Compared with the previous version, users of the new SD-XL only need to use a shorter description to generate images. The human body structure and detail processing of the images are more realistic and more in line with the public's aesthetics. The generated portraits are also clearer and more realistic. .

Web links:

4. DALL-E 2: Master of realistic painting

DALL-E 2 is an image generation and editing tool launched by OpenAI, which is famous for its excellent generation effect and artistic color. The user only needs to input a brief, and it can synthesize the three elements of concept, attribute and style, and generate a realistic image that meets the needs of the user, and at the same time, it can also have the painting styles of different artists.

For example, the user enters three elements: the concept "a puppy", the attribute "on the grass" and the style "Pop artist Andy Warhol style", and it can generate pictures that meet these three conditions. The tool's features also include image editing, style morphing, and more.

Web links:

5. Imagen AI: Generated pictures can be fake

Imagen AI is a text-to-image AI tool developed by Google. It can output portrait photos, oil paintings, CGI renderings and other images according to the user's written prompts. The images have a stronger sense of reality and higher accuracy in language understanding.

Web links:

6. Adobe Firefly: AI drawing + image editing in one stop

Adobe Firefly is an AI drawing tool launched by Adobe and Nvidia. Currently, it has realized the functions of generating pictures from text, converting sketches into pictures, and modifying picture content with one click. It can also modify the automatically generated pictures in layers and output ultra-high resolution rate image.

Web links:

7. One style of writing and heart: support to generate pictures from pictures, and convert pictures to videos

Wenxin Yige is an AI painting product launched by Baidu relying on the flying paddle and Wenxin large model technology. Users only need to enter their own creative text and choose the desired picture style to get a painting generated by Wenxinyige. They can also choose the picture type, picture ratio, and the number of pictures generated at a time. At present, Wenxin Yige has supported more than ten different styles of images such as oil painting, watercolor, animation, and realism.

Web links:

8. Ali Luban Luban: Artifact for E-commerce Mapping

Luban is an image design product independently developed by Alibaba Intelligent Design Lab. Based on AI image generation technology, Luban can complete the design of a large number of Banner pictures, poster pictures and venue pictures in a short time. Users only need to input the style and size they want to achieve, and Luban can replace the time-consuming and labor-intensive design projects such as material analysis, cutout, and color matching manually, and generate multiple sets of design solutions that meet the requirements in real time.

During the "Double 11" promotion in 2017, Luban generated 8,000 posters per second, during which a total of 400 million product posters were produced, which refreshed people's understanding of AI's drawing capabilities.

Web links:

9. Chuangketie AI Painter: You can be a designer even if you don’t know how to draw

Chuangketie, an entrepreneurial design platform, launched an artificial intelligence painting product, AI Painter, and launched two commonly used functional scenarios, "Wen Sheng Tu" and "Tu Sheng Man".

In the "Wen Sheng Tu" scenario, users only need to enter simple required text, select the painting style they want, and the target image can be generated with one click. The existing styles include ancient style, oil painting, color painting, comics, CG, etc.

In the "Picture Man" scenario, users only need to upload the target picture and enter simple text instructions to get a customized hand-painted picture. Its functions also include edge detection, line draft coloring, and pose detection. Function.

Web links:

03. AIGC audio tools: variable sound, cloning, noise reduction

In addition to application scenarios such as text and images, audio is also an application scenario that we have a wide range of contacts in our daily lives. Human voice change, speech synthesis, and cloning in short videos are AIGC's popular research technologies in the audio field, including animation, movies, and character dubbing in games, which can now be done by AI. Technology companies such as Microsoft and Google have also launched their own Text-to-Speech (text-to-speech) services.

AI audio tools can be divided into two types according to different functional attributes: one is sound processing tools represented by So-Vits-Svc, Adobe Podcast AI Voice, Magic Sound Workshop, etc., which use AI technology to repair sound and improve audio quality Or convert timbre, etc.; the second is music production tools represented by MusicLM, Netease Tianyin, Aiva, etc., which can realize the "text-to-music" function in more subdivided fields.

1. So-Vits-Svc: Create the Internet-wide explosive "AI Stefanie Sun"

"AI Stefanie Sun" became popular all over the Internet overnight. Songs such as "Hair Like Snow" and "Rainy Day" "covered" by her have been played more than one million times on Bilibili, and these songs are passed by UP owners. Made by the open source project So-Vits-Svc.

This model uses the SoftVC content encoder to extract the source audio speech features of the real singer, and then transfers it into the VITS speech synthesis model, so that the singer's original voice is preserved. Similar "AI singers" include AI Jay Chou, AI Xu Song, AI Wang Xinling, etc.

In addition to simulating the voices of well-known singers, it can also simulate a large number of real voices based on telephone recordings, video videos and other materials. Previously, some UP owners used this model to communicate with the deceased. However, due to the increasing abuse of the project, the author has removed the project.

2. Adobe Podcast AI Voice: professional podcast audio processing

Adobe Podcast AI Voice is an AI-powered audio enhancer from Adobe that uses AI to improve the quality of blog audio recordings.

After the user logs in to the Adobe account, upload the audio file that needs noise reduction processing, AI will automatically process the audio file, and after the satisfactory audio playback effect is achieved, the user can directly download it to the computer for free use.

Web links:

3. MusicLM: AI model that can sing

MusicLM is a full-true generative AI model released by Google. Through this model, high-fidelity music can be directly generated from text. In addition to text, whether it is humming, singing, percussion, instrument performance, etc., MusicLM can create music based on these existing melodies, and ensure that the music is not distorted.

Its biggest highlight is that it can generate a 5-minute complete track based on one or two prompt words, with various styles, including electronic music, jazz, blues, Pop, etc. The length of the song can also be set in advance, such as a 5-minute complete track or a ten-second humming segment.

In addition, it can also generate pieces played by specific instruments, and even the performance level of the performers can be set. It can also create music according to the characteristics of the times and the place where it is played. Popular music played by an organ by the sea".

MusicLM is trained in a music database of up to 280,000 hours, no matter what style or emotion the song is for it.

Web links:

4. Xunfei Hear: Voice to text anytime, anywhere

Xunfei Hearing is an intelligent voice product of iFLYTEK, relying on iFLYTEK's natural language processing, voiceprint recognition and speech recognition and other voice technologies, iFLYTEK Hearing can meet the voice needs of users in various scenarios, Applicable scenarios cover different occasions such as meeting minutes, lectures, media interviews, and personal writing.

Its advantageous functions also include adding bilingual subtitles to videos, multilingual simultaneous interpretation, and generating subtitles for video conferences, etc., to help users overcome language barriers and facilitate communication and collaboration.

Web links:

5. NetEase Tianyin: Lyrics, music, arrangement and singing are completed in one stop

Netease Tianyin is an AI arranger music creation system produced by Netease, which can create AI music online. Its biggest advantage is that the threshold for music creation is low, and users can complete an original music arrangement according to the guidelines in a short period of time.

Tianyin's workbench includes a number of specific music styles, including pop, folk, electronic, national style, etc. It supports users to create a set of their own chords from scratch, and also supports dragging preset chords into the editing paragraph. Edit the whole song by adding, subtracting, copying, adjusting paragraphs, etc. After all the editing is completed, it will be automatically rendered, and you can get an original arrangement created by yourself after a short wait.

Web links:

6. Magic Sound Workshop: a must-have tool for film and television commentary big V

Moyin Workshop is an AI voice series product launched by AI company Going out to ask. Users can efficiently and conveniently use AI voice technology to simulate a real person's voice with personal characteristics, create AI audio content, and convert text into a real person with one click. voice.

The user quickly imports the article to be synthesized into the sound in the interface, and performs online editing through an operation page similar to the document, so that the document can be converted into audio conveniently. The functions on the editing page include: stress marking, multi-phonetic characters, typo-prone marking, adding background sound, multi-person mixed dubbing, variable speed, rhythm and many other functions.

For users who like technology, finance and other fields, Moyin Workshop has also added AI voice models of CEOs of many related companies in the background, so that users who are familiar with them can use their voices to produce audio content.

Web links:

7. Fake You: Voices can also be faked

FakeYou is a text-to-speech audio editing tool that uses deep forgery technology to generate text-to-speech in different languages and voices. Users can use the voices of their favorite characters to create audio, and it also provides AI text-to-speech functionality.

When the user enters a piece of text to be generated and chooses who wants to read the text, and then clicks the "speak" button, a voice "spoken" by the target person is automatically generated.

Web links:

8. LyricStudio: AI helps you write lyrics

LyricStudio is an online lyrics maker that helps users generate an original lyric that mimics their own style and finds a rhyme for a specific word. Users can upload a text description or musical clip, and it converts it into lyrics that match the content.

According to data from its official website, the tool has collaborated to create more than 1 million songs, and 15% of users on the platform are professional music producers. LyricStudio helped rapper Curtiss King's #1 iTunes album lyrics.

Web links:

9. LALAL.AI: One-click extraction of instrument sounds

LALAL.AI is an online music separation tool that can segment and extract vocals and instruments from music.

Its online music separation technology is entirely based on machine learning and artificial intelligence. Before the previous version, it could only separate human voices. Now it can accurately extract human voices, electric guitars, acoustic guitars, pianos, and drums from audio and video files. , bass and many other instruments.

URL:

10. Aiva: AI Music Producer

Aiva is an AI music tool with the same name self-developed by the AI music company "Aiva". Users can assist musicians to produce and write original music through AI technology. The platform covers a variety of different styles, such as classical, rock, electronic music, pop, national style, Blues, hip hop, etc. On the automatic composition page, there are 11 genres for users to choose from, including Key Signature tune, Time Signature beat, Pacing rate, Instrumentation, Duration, etc.

Aiva has also studied the representative works and music styles of Mozart, Bach, Beethoven and other musicians through deep learning, and established a learning model based on these musical characteristics to help musicians create music. At the same time, Aiva is also the first certified AI composer in history and has published 5 albums.

Web links:

11. Supertone: a voice-changing artifact

Supertone is an AI creative sound studio in South Korea that provides speech synthesis and real-time speech enhancement technology to help users easily create various types of sound content, including simple text reading to works of art, songs, etc., allowing users to change their voices and other ways To alleviate concerns about personal information issues.

Supertone also offers a technology called "VoicePrint," which converts a user's voice into a digital fingerprint that distinguishes it from other users' voices.

Web links:

04. AIGC video tool: automatic editing and generation of storyboard functions are here

Vincent graphs have now become the mainstream AIGC technology, but text-to-video generation is still in its infancy.

New York-based AI startup Runway has developed a generative video model Gen-2, which can generate a highly composite video from a simple description. Other companies have also joined in, such as Text2Video-Zero, Video-P2P, and TemporalNet launched by the image editing platform PiscArt, and Text-to-video developed by Ali. Text-generated video may also enter fierce competition in the near future. stage.

1. Deepfakes: AI video face change

Deepfakes are now synonymous with AI-synthesized videos. Microsoft launched FaceShifter, which can process a blurry original image into a clear and credible forged picture; Disney and ETH Zurich jointly developed and launched a megapixel-level Deepfakes video production tool, and in the "Star Wars" series In the movie, Deepfakes were used to bring deceased actors back to the big screen.

However, the security risks caused by this have also come one after another. For some high-definition and extremely natural light videos, even the most sophisticated Deppfakes algorithm cannot accurately identify them.

As a result, as early as the 2020 U.S. election, Facebook announced a complete ban on the use of Deepfakes on the platform, and YouTube and TikTok were no exception, explicitly prohibiting the illegal use of Deepfakes technology in videos. The "Civil Code" that will be implemented in my country in 2021 also points out that major video platforms need to strictly restrict the content of AI face-changing videos, and they must not be used at will without permission.

2. Runway Gen-2: Generate blockbuster movies in 30 seconds

Gen-2 is an end-to-end Transformer model launched by the start-up company Runway. Users can use pictures and text as conditions to generate an original slow-style video from scratch.

The video resolution it generates is as high as 1280×720, and the duration is about 30-60 seconds. Currently, the following functions can be realized: generating video, generating images, expanding images without limit according to text prompts, mixing image styles, training AI models, Remove an element in the video, subtract the background, etc.

Recently, Runway launched its first mobile application, using the Gen-1 model, users can upload text, pictures or videos on the mobile phone, and let the model transform the style of the video according to the content.

Web links:

3. Make-A-Video: convert text to video directly

In September 2022, Meta launched its own text-to-video software "Make-A-Video". After the user enters a few simple word descriptions, the software will create a silent video.

In the official demonstration video, the user can get a video of a few seconds by entering text descriptions such as "a young couple walking in the heavy rain" and "a teddy bear who has been painting a portrait". In addition, Make-A-Video can also animate static pictures, which is based on the "Vincent diagram" technology.

According to the official, the model is trained using image synthesis data and unlabeled videos. After learning, the model can "predict" what will happen next to the image, where it will move, and move to where the image will be in a very short time. The location where it appears to form a short video.

Web links:

4. Shangtang Zhiying: short video expert treasure artifact

SenseTime has launched a one-stop advertising and marketing platform for SenseTime, which includes the short video creation engine "SenseTime", which can generate creative short videos with one click, including script generation, background replacement, horizontal and vertical screen replacement, and subtitle generation. A variety of services for video advertising production can help advertisers save the cost of advertising content production.

The "Video Element Analysis" service included in SenseTime can analyze and extract information such as the length, scene, scene, character, props, and lines of each shot in a short video through AI video structuring technology, and automatically create A shot script greatly reduces script writing time and effectively assists creators in secondary creation.

In addition, the platform also provides a large number of popular video scripts to provide creators with creative inspiration.

Web links:

5. Decoherence: Generate video with one click of the picture

Decoherence is a tool for creating AI videos where users can choose from a variety of AI styles.

Web links:

6. Tencent Zhiying: short video creation artifact

"Tencent Zhiying" is mainly aimed at short video creators, and its featured functions are genuine copyright materials and digital human broadcasts. Users can generate a digital human video by uploading photos and text. Users can also use it with the intelligent AI dubbing function to choose different timbres for digital humans.

Web links:

05. AIGC office tools: AI+OA realizes one-click "from scratch"

On March 17, Microsoft officially released Microsoft 365 Copilot, which integrates the capabilities of GPT-4 and ChatGPT into Office tools, and launched the Business Chat function integrating Office 365 data, which improves the level of digital office and saves employees from inefficient , Liberated from repetitive labor.

Kingsoft Office, as a leading company in the domestic collaborative office field, also launched a generative office platform "WPS AI" with large language model capabilities in just one month, becoming the first ChatGPT-like application in the domestic collaborative office track. In addition, companies such as Baidu, ByteDance, and DingTalk have successively launched their own AI collaborative office tools.

The OA (Office Automation) application system has gradually developed and matured. As a bridge connecting employees and enterprises, it may become the entrance of the big language model in the B-end ecology in the future.

In addition to the field of collaborative office, AI tools can also be applied in more vertical scenarios and combined with more practical needs. For example, "AI + language learning" has DuolinguoMax, and "AI + e-commerce delivery" has created a smart version of e-commerce. eCommerce website Shopify, e-commerce marketing tool eCommerce s, etc. "AI+ programming" makes GitHub Copilot X a powerful assistant for developers, and "AI+ mind map" has Chatmind, which can generate a mind map with a sentence description .

1. Microsoft 365 Copilot: Gpt-4 version of Microsoft Family Bucket

Microsoft 365 Coplilot followed GPT-4 in the early morning of March 17, and all office software including Word, Excel, Powerpoint, Outlook, Teams, etc. were launched with generative AI functions.

In Word, Copilot only needs a simple prompt to create a first draft, and can also adjust the tone of the article according to the user's needs, such as professional and serious, enthusiastic and casual, etc., and can automatically delete the same place in the article , for further simplification.

Copillot in Excel can help users analyze data, directly analyze data trends and visualize data analysis results.

What's even more amazing is that Powerpoint can already directly generate a PPT, and Copilot can directly convert an existing document into a PPT with marked sources. If users feel that the PPT is too lengthy, they can directly use the text description to compress, adjust the layout or format the text with one click.

Copilot in Outlook can help users classify emails according to certain criteria, summarize and refine the subject of long emails, and transcribe several keywords or drafts into official emails.

Web links:

2. Google Workspace: technology + office = artifact for workers

Google Workspace is a Google workbench that includes office tools such as Docs, Slides, Sheet, and Gmail. Google announced in March that it will integrate AI into these tool components. After accessing the generative AI model, users will be able to create a complete email, business plan, or advertising marketing fee sheet with the help of these tools by entering a short text description.

In Docs, generative AI can help users draft the first draft of text, polish and revise text, proofread and correct errors; Gmail can reply and summarize emails, mark important matters, etc.; Slides can automatically generate images, audio and video according to the theme and insert them into the template; Sheet can automatically perform data processing, table sorting, context classification, and even raw data analysis.

At present, Google adopts a flexible payment plan, which is divided into basic business novice version, business standard version and business Plus version, allowing users to subscribe according to their actual needs.

Web links:

3. Baidu Ruliu: AI + knowledge management

Baidu Ruliu launched the "Ruliu Intelligent Work Platform 2.0" for the enterprise service market at the end of last year, including three intelligent product matrices: intelligent knowledge management, intelligent conference, and intelligent workbench.

In intelligent knowledge management, there are three knowledge management applications of "intelligent knowledge base", "search and recommendation dual engine" and "knowledge star chain", which gather scattered documents, emails, notes and other files in one place, Employees can find the required documents and knowledge in the most convenient way.

Smart meetings intelligently connect employees, spaces and equipment. Before the meeting, Ruliu Conference Assistant can help you check the schedule of participants, reserve the best meeting time, and send the meeting materials; during the meeting, Ruliu Assistant can record the speeches of the participants and convert the voice into text in real time, marking key information; After that, the meeting assistant will automatically generate a meeting to-do, which is convenient for employees to review the key points of the meeting.

Different work cards are collected in the smart workbench to make the task system more flexible and clear, and different work cards are matched according to employees in different positions. For example, the HR workbench is embedded with job cards for interview assistants and recruitment management modules; the manager workbench includes three-dimensional work cards for efficiency tracking, collaborative analysis, and process management, providing managers with team task data.

Web links:

4. Dingding slash "/": Magic wand generates applets with one click

A week after the large-scale model of Tongyi Qianwen was released, DingTalk announced its official access. After accessing the Qianwen large model, users only need to use a "/" slash to perform intelligent office work. The main usage scenarios include group chat, documents, video conferencing and applet development.

In the group chat, new entrants only need to enter "/" in the dialog box to get the contextual points of the group chat, and the slash can also generate to-do items, schedule appointments, and make emoticons for important meetings.

In a one-to-one chat, users can directly use slashes to create a chatbot to let it automatically learn knowledge and answer questions.

In documents, slash can automatically generate copywriting and posters in various styles; in video conferences, slash can summarize meeting points, to-do items, etc. with one click.

In addition, the most unexpected function of slash is to generate small programs in natural language and use them in the group in the form of "Dingding cool application".

5. Kingsoft Office WPS AI: AI writes documents

On the same day that DingTalk announced the access to the large model and the launch of the smart office assistant "/", Kingsoft Office, a leading domestic office software company, also officially announced the launch of "WPS AI". The underlying large model is provided by MiniMax, which currently includes multiple functions such as content generation, multi-round dialogue, and content optimization. In the future, it may evolve into the domestic version of "Microsoft 365 Coplilot" and be fully embedded in the WPS suite.

Web links:

6. Feishu My AI: Bytedance version administrative assistant

On April 11, Feishu, the office platform of ByteDance, also launched the intelligent AI assistant "My AI". Its functions include automatically summarizing meeting minutes, creating reports, continuing to write and optimize text content, etc. In Feishu, My AI can also help users create schedules and search the company's internal knowledge base through dialogue. However, My AI is still in progress, and the public beta and launch time have not yet been announced.

Web links:

7. Shopify: AI transforms e-commerce customer service in seconds

On March 1, after ChatGPT announced the opening of API, the cross-border e-commerce service platform Shopify took the lead in integrating. After integrating ChatGPT, Shopify can use intelligent customer service to communicate with users, help consumers make personalized recommendations, and save purchase time; ChatGPT also conducts review data analysis, title and keyword optimization, marketing copywriting, and intelligent website development programming for platform products and many other functions to help improve the operating efficiency of e-commerce websites and optimize consumer purchasing experience.

Web links:

8. eCommerce ChatGPTs: eCommerce tipster

Web links:

9. GitHub Copilot X: Programming Xiaobai can also write code

After Microsoft launched the new version of Bing search engine, Edge browser and Office family bucket, its code hosting platform GitHub also released Copilot X, which introduced ChatGPT into the integrated development environment, even users with zero code foundation can rely on "move your mouth" Write out the code.

In GitHub Copilot Chat, users can send it instructions to write code in a chat window. For those codes that run abnormally, it can directly find the bug (error) and modify it; in GitHub Copilot Voice, it can even be implemented. From voice to code in one step, the developer directly speaks and gives natural language instructions, and it can generate relevant codes.

In addition to the above functions, if the user does not understand a certain line of code, just let Copilot explain the function of the code in Chat.

Web links:

10. Fireflies: meeting minutes easily resolved

Web links:

11. Feishu Miaoji: A thousand words to text, a word is worth a thousand words

Web links:

06. AIGC life tools: cooking, taking notes, making travel guides

Let AI become the steward of life

In addition to highly applicable AI tools such as text generation and audio and video editing, various new AIGC products have emerged in daily life.

For example, ChefGPT helps generate recipes for users who have a headache every day, Dover Autopilot, an AI recruitment tool that provides high-quality talent resumes for headhunters, BibiGPT, which can take notes while watching videos, and Bedtime Story AI, which can generate short stories before going to bed. WatchNow, which recommends movie lists for personal preference, etc., fully intelligent life is no longer a plot only in science fiction movies, and AI has fully entered our daily life.

1. BibiGPT: a friendly tool for college students, enabling one-click transfer of videos to notes

The video is too long to summarize the key points? Too rushed to take notes while watching a video? BibiGPT, an audio and video summary software developed based on ChatGPT, solves these problems well. For videos on Bilibili and Youtube, BibiGPT can summarize the key content with one click. Users only need to paste the video link they visited on the search box and click " "One-click summary", you can get a video summary note.

Web links:

2. Dover Autopilot: AI recruitment software

Dover Autopilot is an automated recruitment tool. Recruiters only need to enter a simple job description link, and it can find job seekers that match job requirements within minutes through data sources such as LinkedIn and other job search websites. It can also automatically generate Personalized emails for candidates.

Web link: dover.com/start

3. ChefGPT: A Recipe Encyclopedia for Food Novices

This is an AI recipe recommendation tool. When the user enters the existing ingredients and tools at hand, as well as the reserved cooking time, it can recommend a recipe that meets the needs.

There are two modes in the page, one is gourmet mode, and the other is select all mode. The gourmet mode has higher requirements for user autonomy, requiring users to select ingredients and kitchen utensils before inputting them into the webpage, while the select-all mode is more friendly to "kitchen noobs", and can get a copy that meets the requirements without using their brains. Recipes for real needs.

Web links:

4. Journeai: Travel Guide for Backpackers

Journeai is an AI-based chat travel advisor, which aims to create personalized travel routes for users. It can generate itineraries according to user preferences, including activity arrangements and travel partners. explore.

This tool is not only suitable for vacationers who need to customize personalized itinerary arrangements, but also suitable for assisting travel agencies to improve user experience.

Web links:

07.

AIGC Content Detection Tool: Counterfeit AIGC

Leading the way in content identification

With the development of AIGC technology stepping into the fast lane, it has also caused a lot of false information, text plagiarism, academic fraud, copyright disputes and other adverse effects and related negative events. Unknown security disputes are unavoidable, which requires the development of relevant technologies for detection and screening.

Although there are not many AIGC content detection tools currently on the market, they can already accurately distinguish AIGC content such as generated text, pictures, and sounds. Plagiarism detection system CrossCheck etc.

1、Copyleaks：

Copyleaks is currently one of the most popular AI content detection tools in the world. The most prominent advantage is that in addition to detecting English content, it can also detect content written in Spanish, French and other languages. For texts that are all generated by AI, Copyleaks can achieve 99.99% recognition accuracy, but for text content that is half true or false, it will mark it as artificially generated text.

Web links:

2. AIGC-X: Identify the authenticity of Chinese text with sharp eyes

People.cn Information Technology Company, a subsidiary of the People’s Daily, which focuses on content risk control, has developed the first AI-generated content detection tool in China, AIGC-X. This tool can be used to distinguish machine-generated text from artificially It can detect and screen fake news, content plagiarism, spam, etc., and especially provide technical support in gray areas such as false information, academic fraud, and phishing.

However, AIGC-X currently only supports the detection of Chinese content, and the detection ability of images, audio and video content needs to be improved.

Web links:

3. DetectGPT: anti-reconnaissance tool, fraud and cheating are inevitable

The zero-sample detection tool DetectGPT was developed by a research team at Stanford University in the United States. It is mainly used to combat the phenomenon of paper generation that is common in universities. A research paper entitled "DetectGPT: Zero-Shot Machine-Generated Text Detection using Probability Curvature" has been published on the preprint website arXiv.

This detection tool proposes a new indicator for judging the text generated by the large language model. It only needs to scan the content uploaded to the web page to determine whether the content in the text is generated using the large language model.

Web links:

4. NetEase NetShield: Accurate detection of sensitive words

NetEase NetShield, based on NetEase's years of experience in the industry, provides personalized matching models and customized detection solutions for the characteristics of text spam. Content diverted for third parties will also be automatically filtered out.

Web links:

5. Sumei intelligent text detection: rapid identification of risky text

Sumei uses a full-stack intelligent content recognition engine to effectively identify sensitive, prohibited, pornographic, violent, abusive, advertising diversion and other risky text content in various scenarios, helping users further identify risky information. At present, Sumei has been able to automatically detect 175 overseas languages, and supports risk label identification in 18 mainstream languages such as English, Arabic, Thai, and Indonesian.

Web links:

08. Conclusion: AIGC set off a productivity revolution in all fields

Become a Copilot for Creators

The AIGC track is crowded. In addition to writing, image generation, audio and video editing, office assistants, content detection, etc., there are more subdivisions waiting to be explored.

Today, AI is striding into the field of digital content production. In addition to being comparable to professionals in writing, question and answer, painting, and century-old cities, it has also demonstrated the powerful understanding ability of large language models. But it can only serve as a "Copilot (co-pilot/assistant)" to assist humans in making decisions, assist creators in continuous production and iterative ideas, and will not replace those truly valuable work.