Image source: Generated by Unbounded AI

Looking back on 2024, it will undoubtedly be the year when AI applications explode.

An endless stream of application products, large and small, are emerging, and they are becoming more and more eye-catching. There are even dozens or even dozens of products with the same function.

At the end of the year, we launched a heated discussion on who is the most powerful AI assistant for creators?

In the end, we found that different products seemed to only be able to solve a single point of work. If we want to solve all workflows on one product, the answer is: no.

Really?

At the end of 2024, the computer version of Doubao "Qing Mimi" will undergo a major update, changing the attributes of Doubao from a conversation-based product to an AI browser, Compared with the previous fragmented workflow, if all links can be completed in one product, the efficiency of the article editor and the All in one product generated by AIGC will undoubtedly be greatly improved.

Today, we challenge whether we can complete efficient creation using only bean bags without switching to any product.

1. Make a different AI browser

It’s hard After three nights, dozens of web pages were opened. There are all kinds of information on the Internet. To find valuable information efficiently in the ocean of information, creators all know how deep the knowledge is.

When we habitually open the browser to search, we find that Doubao not only supports AI search, but also transforms into a browser. We can directly search for official website information, financial reports, papers and other primary sources. . When I imported the bookmarks in the original browser with one click and switched seamlessly, the usage experience was almost the same as Chrome.

When AI and search began to be combined, AI search evolved into two forms in the past two years: first, LLM itself has search capabilities and generates accurate answers by searching for real-time information during conversations; second, It is a type of independent product represented by Perplexity that is specifically responsible for AI search and cannot directly generate content.

Of course, these two forms are relatively convenient and inconvenient for users, but they are often only suitable for retrieval of second-hand information. When searching for tools, services, or primary sources, you still need to open the browser , cannot solve all problems, which also leads to the fact that creators still need to supplement and confirm information through traditional search methods after AI search.

In the PC version of Doubao, it not only combines the above two forms, but also takes into account the real needs in the usage scenario and retains the native capabilities of the browser.

What is surprising is that although the functions are extremely complex, Doubao still maintains a simple page through its product capabilities.Convenient and simple operation, even a novice can operate it directly.

When we ask an accurate question that needs to be efficiently integrated and get the answer directly, we can choose to directly "Ask Doubao" and enter the dialogue page.

But sometimes, we also need to organize the existing information to form a link for thinking. At this time, we can directly select AI search, and at the same time, in the AI ​​search, click "Return to original" in the upper right corner Web page" will jump back to the Baidu search page.

But the function of the Doubao PC browser is not just to build the browser itself. More importantly, with the browser as a carrier, AI assistance can be more conveniently invoked during browsing.

For example, in the process of researching AI technology, creators often search and read many papers, but to laymen, they always look at one and the other, and the translation and understanding of many nouns are not accurate. Equipped with an academic search function, you can search AI and translate the original text at any time during browsing.

I understand the text, but when it comes to graphic demonstrations or mathematical problems, I have difficulties again.

The Doubao computer version of the browser is equipped with a "screenshot search" function. When a certain part of the image is not understandable, you can directly take a screenshot and use Doubao to analyze its meaning.

If you want to browse a paper quickly and efficiently, you can also summarize the web page and directly generate a brain map to help us quickly understand the complex information in the paper. At the same time, you can play audio based on the web page content to expand its use. scene.

In addition, Doubao also takes into account that in addition to text content, creators often also need to browse a large number of more complex content forms such as audio blogs and videos. The computer version of Doubao has upgraded its browser function to understand pictures. and visual comprehension of videos.

In the past two weeks, the entire technology circle has been following OpenAI’s real-time broadcast, and I can’t wait to watch it frame by frame and study it repeatedly. Watch videos through the browser function of Doubao PC version, you can understand and summarize the video, and you can even directly locate the video content by searching for keywords. Don’t be afraid even if you miss it for a few days, you can also accurately search for “Sam Altman’s original words” during research ” or “What updates are there about o1?” You no longer have to rely on memory to find videos frame by frame.

Searching, browsing, translating, organizing and summarizing, in the new Doubao PC browser function, accessing information has become a simpler thing. What makes us even more impressed is that compared to clicking a bunch of plug-ins in Chrome, the Doubao PC browser function seems to understand "my" needs better and is more convenient.

2. Check and correct while writing, so that the computer screen is not so busy that it smokes

< /p>

In the past, searching and writing had to be done in different products to extract effective information from web pages, organize and process it.

This results in the computer screen often being divided into twoblock, or install multiple screens.

And this is not just a matter of efficiency. Copying and pasting on both sides, and switching back and forth between the two pages often cause bugs and format errors, which affect the accuracy of the creative content.

I believe every creator has experienced such "pain". After the arrival of AI, we thought it would improve efficiency and found that there was another modification step: checking information - writing a first draft of a large model - copying and pasting in the editor - manual modification.

This process creates a longer chain, and the invisible efficiency loss in the middle is very large.

Then the question is, can it be generated while checking, and can it be modified at the same time?

The Doubao PC version has launched an integrated "conversation + editing" editor. The page form is similar to ChatGPT Canvas, but its use is more in line with the creative habits of creators, and the product functions, AI generation and creators are refined. The organic integration of modifications opens up a new way of human-machine writing.

Silicon Star has pre-tested the editor function of Doubao. When the number of writing instructions entered in the dialogue portal or the Help Me Writing portal exceeds 300 words, the editor can be automatically triggered, seamlessly and smoothly switching to writing mode. , can solve almost all problems on one page.

The left side is a free dialogue page. In the dialog box on the left, frame requirements are proposed. On the right side, a more detailed framework is generated, and multiple rounds of dialogue are used to improve the content of the framework, which is divided into points, levels, and clear and meaningful. The sequence shows the logic and structure of the article.

After the beanbao generates the first draft, the creator can re-create and arrange the generated content. During the creation, if you are not satisfied with a certain paragraph, you can directly propose modifications in the conversation and make partial adjustments. Modify and generate directly in the article, unlike traditional AI, which will not regenerate a completely new article, which will produce excessive text and reduce the final writing efficiency.

The requirements are put forward on the left, and the results are generated on the right

Of course, in the editor, creators can also rewrite and modify themselves with AI at any time, supporting the expansion of the full text or specified paragraphs and sentences. / polish / abbreviation and other AI rewriting functions, you can also customize prompt words, modify only the target paragraph or sentence, and make independent replacements.

For example, the generated result of the first paragraph of Doubao is slightly colloquial. When we need a more serious business report, we can adjust it to a more written text through Prompt, and we can freely choose to replace the entire paragraph or part of it. Improved editing flexibility.

The picture above is before rewriting, and the picture below is after rewriting

We found that each requirement beanbag editor can generate an independent document cloud storage for easy retrieval later.

Finally, we can use the full-text optimization function to retouch the language of the article that has just been modified in sections to ensure the integrity and consistency of the article.

The picture shows the generated results after polishing the full text

When it goes beyond the scope of the article, such as writing research reports and papers, a stronger logical structure andRelevance, the step-by-step writing function of Doubao supports the writing of very long texts of more than 10,000 words, and supports three-level headings. It breaks down complex tasks into a structured structure, and the creator can modify each part of the outline in detail until they are satisfied.

In step-by-step writing, AI writing can also become more and more "like you". Send the author's past works to Doubao, and Doubao can imitate the language style and article tonality, making it more personalized. ized production.

These seemingly independent capabilities actually serve the same goal: ensuring the high availability of generated content. From assisting creation to directly generating usable results, AI is reshaping a new way of human-computer interaction.

From the perspective of product form, the page layout of Doubao Editor is similar to that of the Canvas Editor launched by OpenAI in October. Both of them chose to integrate the "dialogue" and "editing" functions. Construct a new human-machine collaboration paradigm to try to open an era of co-creation between AI and humans.

3. Sensible AI editor, you don’t even have to find pictures by yourself

In the process of graphic creation, images are an indispensable and important component. But for most text workers, although writing text content is relatively easy, finding suitable picture materials is often difficult.

Especially when the picture is required to be consistent with the context and does not involve copyright issues, it is even more difficult to find suitable materials.

In the past, we often used Vincent diagram tools for secondary creation, but this required rewriting prompts, and sometimes garbled characters were generated due to unfamiliarity with professional terminology, making the generated images impossible to use in articles.

The sensible Doubao editor can even connect the accompanying pictures with contextual semantics and mastery of the entire article to understand and generate more suitable illustration content without even raising a prompt.

In the editor, click ➕ directly to invoke the entire network image search or AI image generation capabilities without even proposing new prompt words. According to the contextual semantics, for the "Inspirational Creative Workshop" Metaphorically, a magic prompt board is generated. If you are not satisfied with the generated results, you can regenerate new images.

Of course, we can also directly propose new prompt words in the conversation to describe the picture information more accurately. As shown in the picture above, Beanbao generated "AI Search: Information Highway". Based on this idea, we proposed a simple picture idea, and the illustration function of Random Beanbao generated four illustration-style illustrations.

I believe you who are sharp-eyed have also noticed that, yes, in the PC version of Doubao, the previous "controllable Chinese generation ability" has also been updated. Doubao accurately generated highways with information characters, and even placed them in different locations in the four pictures.

However, for images that are mainly text supplemented byFor creators, it is often the norm to re-create images. For example, in the picture below, it is obvious that garbled characters appear in the billboard. We used the "region redrawing" capability and proposed a new Prompt to turn the billboard into Tree, "erase" and "expand" three different modification methods, the following image is edited in real time to ensure usable effects.

Of course, based on the two relatively basic content forms of pictures and text, it is reported that the computer version of Doubao will also launch PPT generation capabilities, and the Tiancai ability of workers will finally be launched!

In addition, the much-anticipated video generation capability is also in internal testing. You can also look forward to the code capability that OpenAI has just updated in Canvas.

At this point, Doubao really integrates searching, writing, generating pictures, and videos. I wonder if you think this challenge is successful?

4. Ending

In the Neolithic Age, humans first made wood Tools such as axes, awls, and flints were integrated and used together, giving birth to the first tool revolution in history. Today, the AI-driven All in One platform emerges, integrating multiple functions into one, bringing profound changes to knowledge creation.

Different from the previous state of tool fragmentation, the All in One platform is reshaping the way creators think and create, opening up a new creative paradigm. Creators are no longer limited by tools. In turn, tool sets are reshaping human cognitive methods. This not only improves creative efficiency, but more importantly, gives birth to a creative model of deep human-computer interaction.

From another perspective, in a new era of AI explosion, users are "voting with their feet". The reason behind All in One is that users are willing to choose Doubao in different scenarios. Natural results. On the one hand, multi-modal input and multi-terminal deployment make Doubao ubiquitous around users. QQ Doubao is the fastest and most timely.

On the other hand, QQDoubao is the “fastest” not only in terms of speed, but also because it is effective and provides answers in the simplest and most intuitive way. You no longer need to spend time reading books or searching online, you only need to ask questions, especially since the computer version of Doubao integrates almost all capabilities, you can really ask Doubao about anything.

The seamless integration of the All in One platform marks the advancement of AI-assisted creation to a mature stage. It brings together human intelligence and AI capabilities to bring new large-scale improvements to knowledge productivity. This is not just a simple function integration, but also the only way for AI assistance to mature.