From Real Materials to WeChat Official Account Articles and Xiaohongshu Image-Text Posts

An image-text workbench for content creators: collect information, research topics, write and review articles, create branded visuals, then organize content into finished pieces suitable for different platforms.

Consistent creation is difficult because of a chain of trivial handoffs: information is scattered across different websites, and when writing you cannot find the original links; once a draft exists, facts still need to be checked one by one; after the article is finalized, you again need to make a cover, body images, and Xiaohongshu cards. Content Studio connects these tasks, so a topic stays in the same creative space from source intake and research writing to visual production.

It does not equate “generated successfully” with “ready to publish.” Sources are traceable, articles are editable, images can be checked one by one, and ultimately the creator decides whether content enters the publishing workflow.

What the project can do

Module	Main functions	Output
Workbench	Enter creative ideas, continue recent drafts, view publishing queue	Creative tasks and article entry points
Topic Planning	RSS / Atom subscriptions, information filtering, public article collection, source library, topic Brief	Reusable materials and topics
Research Creation	Retrieve materials based on natural-language topics, organize evidence, write, record process, review and revise	Article drafts and versions with sources
Content Creation	Rich-text editing, AI-assisted writing, cover and tags, layout preview, publishing checks	Editable article drafts
Article Images	Brand visual settings, cover and body image planning, online image generation, one-by-one verification and insertion	Article covers, body images
Xiaohongshu Cards	Generate card scripts from articles, edit and preview card by card, export images and note copy	A set of image-text cards and accompanying copy
Content Assets	Find saved covers by article title, return to the original article	Reusable article covers
Publishing Center	Manage articles to publish, select WeChat Official Account, submit to WeChat Official Account draft box	Drafts pending handling in the WeChat Official Account backend

Below, these features are introduced in the order of one complete creation process.

01 | Starting from information input: subscriptions, collection, and topic selection

“Topic Planning” provides two entry points for materials.

The first is subscribing to information. Creators can add real RSS or Atom addresses, set categories for feeds, manually refresh or unsubscribe; the page also provides feed suggestions that can be added directly. The information list supports viewing by category, loading more, and opening details for a single item. In the details, you can view the source, publish time, original link, and the summary or body provided by the feed. For subscriptions that only provide summaries, the interface clearly prompts you to open the original to read the full text.

The second is collecting public articles. After pasting a public webpage URL, the system attempts to extract article information and save it to the current user’s source library. For pages that cannot be read automatically, you can manually add the title, source name, link, summary, and notes. The source library supports search, making it easy to find saved articles later. Collection only targets public pages and does not bypass logins, CAPTCHAs, or paywalls.

After materials enter the workbench, you can select 1–8 collected articles to generate an original topic Brief and clarify the writing direction; you can also select 1–8 RSS information items to generate an article draft with original links. The Brief can be saved, edited, and then carried into the next step as a writing reference. Here, materials play the role of “research basis”; the final viewpoints and facts still require the creator’s judgment.

Figure 1 | Manage feeds, browse information, and start writing from real materials on the Topic Planning page.

02 | From idea to first draft: reviewable research creation

When you have a clear topic, you can directly describe the theme you want to write about, the target readers, and the questions you hope to answer on the workbench or the “Research Creation” page. The Research Creation page can also provide topic suggestions based on subscribed RSS information. Before starting, you can set the material scope (whole web, my RSS, or both), article length, writing style, and add public source links.

Figure 2 | Choose a direction from RSS recommendations, or directly enter your own topic, and start researching and writing.

After submission, the system creates an independent creation task. The task goes through stages such as planning, retrieval, reading, evidence organization, outlining, writing, checking, and saving; the page displays the actual status and event records. The task list can return to recent creation tasks. During execution, you can add requirements or links, and cancel the task if necessary; if it fails, lacks materials, or needs changes after review, you can research again or write another version based on saved context.

A research task not only saves the article, but also provides several checkable views:

• Research Materials: view source title, site, original link, publish time, fetch time, and whether the currently saved content is the body or a summary; unsuitable sources can be excluded in the next version, or restored.
• Evidence: view the specific statements used in the article, supporting excerpts, and corresponding original sources, making it easy to return to materials for verification.
• Article and Review: read the generated article, open the editor to continue editing; users with review permission can approve the current version, or specify revision comments and request changes.
• Version Record: keep revised versions, and support side-by-side viewing of historical versions and the current version.
• Cover, Images, and Layout Preview: continue from the same task into visual production, and preview the article theme.

This makes the creation process a reviewable record: what the topic was, which materials were referenced, and how the article was revised can all be viewed along the task. The evidence organized by the system is a verification clue, and cannot replace human review of facts, citations, and high-risk statements.

Figure 3 | The left side records dialogue and execution stages; the right side brings together the article, materials, evidence, and versions.

03 | From first draft to work: editing, layout, and publishing checks

“Content Creation” can create a new blank article or continue editing drafts generated from research tasks or RSS information. The editor supports headings, body text, bold, italic, strikethrough, code, H1 and H2 headings, lists, quotes, image insertion, as well as undo and redo. While writing, you can also open “AI Collaboration” and describe facts, examples, or paragraphs you want to add; the generated content is inserted at the end of the body for the author to continue adjusting.

On the right side of the article is the publishing preparation area. Here you can set or change the cover, jump to the brand image and Xiaohongshu card workbench, select content tags and the target WeChat Official Account, and switch the article layout theme. In addition to the default layout, the project provides three themes: Editorial Lime, Checklist Signal, Night Reading Archive. Creators can preview the current theme first and check the cover, body images, and layout effect.

The editor also provides a set of clear publishing checks: whether the title is 8–64 characters, whether a cover is set, whether the body reaches 300 characters, whether the body contains images, and whether content tags are added. Among these, the title, cover, and body length are required to enter the publishing queue; images and tags count toward the check score but are not required. This set of checks focuses on content structure and does not mean the article’s facts have been verified. Drafts can be saved at any time; once ready, choose to add them to the publishing queue.

Figure 4 | Body editing, cover, layout theme, publishing target, and checks are centralized on the same page.

04 | Create a complete set of covers and body images for an article

“Article Images” starts from a saved article, first plans the cover and body images, then generates the finished products. You can customize the brand name, the overall visual style prompt, the default image aspect ratio, and the number of body images; each image can still individually modify its title, subtitle, scene prompt, aspect ratio, and text display method. Text can be set to not display, display only the title, or display both title and subtitle; the brand name on the cover can also be turned off separately.

There are two ways to generate images:

1. Pure AI image generation: an online model directly generates the finished product containing both visuals and text.
2. AI text-free base image + browser typesetting: an online model first generates the visuals, then text such as the title is typeset in the browser.

Creators can generate the whole set at once, or generate or regenerate image by image. Each image shows its corresponding article basis, making it easy to check whether the visuals deviate from the content. After generation, each image must be confirmed one by one: in pure AI mode, check Chinese characters, numbers, and the brand name; in local typesetting mode, check whether the base image contains mixed stray text. Only verified images can be inserted into the article, and the first image can update the article cover. Finished products can be downloaded individually, or packaged and exported together with the image list; the image plan is saved for future modifications.

05 | Adapt a long article to Xiaohongshu image-text posts

“Xiaohongshu Cards” also starts from a saved article, but the output target becomes a set of image-text cards suitable for mobile reading. The system first organizes the card script, then creates the cover, 2–10 information cards, and an optional ending card. Cards can choose a 3:4 or 1:1 aspect ratio and unify the brand visual; information cards can also display sequence numbers as needed.

Each card can individually adjust layout, text hierarchy, title, subtitle, body text, scene prompt, and order. Text hierarchy includes pure image, title only, title plus subtitle, and a complete information card with body text. The generation method is the same as Article Images: choose pure AI image generation, or text-free visuals with local browser typesetting. The workbench previews finished products in mobile aspect ratio and displays the card’s factual basis and verification status.

In addition to images, the page also provides editing for note title, body, and topic tags. After completing card-by-card proofreading, you can export sequentially named PNGs, note copy, and a fact-checking list. What is provided here is an export of publishable materials; it will not automatically post to a Xiaohongshu account.

Figure 5 | The left side previews the whole card set; the right side edits layout, copy, and visuals card by card.

06 | Content assets, WeChat Official Account connection, and Publishing Center

“Content Assets” currently centrally displays saved article covers in the current space. You can search covers by article title and return from a cover to the corresponding article to continue editing. It gives visually used results a browsable entry point.

WeChat Official Accounts are managed through “Workspace Settings → WeChat Official Account Connection” to enter the account management page. One workspace can connect multiple WeChat Official Accounts, and when editing an article you need to explicitly choose the publishing target. The currently available method is to fill in the WeChat Official Account name, AppID, and AppSecret, with Access Token optional; the list shows account connection status, and accounts can be removed. The interface retains a QR-code authorization entry, but the required WeChat third-party platform authorization process is not yet complete, so manual connection should be used currently.

“Publishing Center” centrally displays the article status and publishing queue of the current space. After an article is ready, a user with permission can submit it to the draft box of a connected WeChat Official Account. For articles generated by Research Creation, the current revision also needs a record of passing human review before it can be submitted. Submitting to the draft box does not equal mass sending from the WeChat Official Account; for formal mass sending, you should still operate in the WeChat Official Account backend.

07 | Auxiliary capabilities in daily use

• Personal creation space and login: content, sources, drafts, and images are saved around the current user space; the interface provides login and logout entries.
• Quick navigation: the sidebar centrally displays eight main modules; the top search box and ⌘ K can quickly jump to pages and creation tools. Currently, quick search is mainly used for navigation.
• Appearance switching: choose light, dark, or automatic mode that follows the system.
• Writing model settings: configure a DeepSeek or OpenAI-compatible interface, fill in the model name, HTTPS endpoint address, and key, and test the connection before saving; keys are encrypted and not displayed back in the settings interface.
• Responsive workbench: the desktop version is presented with a sidebar and main workspace; narrower screens use an openable navigation menu. The card workbench provides mobile aspect ratio previews.
• Real errors and empty states: when there is no subscribed information, drafts, covers, or available sources, the page displays the corresponding state; when external search, model, and image generation services fail, errors are shown, and sample content is not used to pretend generation succeeded.

A complete creation can proceed like this

1. In “Topic Planning,” subscribe to industry RSS or collect several public articles, saving reviewable materials.
2. Select information to generate a Brief, or directly describe the topic in “Research Creation,” and specify the material scope and writing style.
3. Review the sources, evidence, and first draft in the task; if necessary, add links, exclude sources, revise the article, and complete review.
4. In “Content Creation,” adjust language and structure, and choose the cover, layout theme, tags, and target WeChat Official Account.
5. Go to “Article Images” to generate and verify the cover and body images, then return to the article to check the reading effect.
6. If Xiaohongshu image-text posts are needed, continue breaking the article into cards, proofread card by card, and export images and note copy.
7. After the article passes publishing checks and any required review, enter the Publishing Center and submit it to the WeChat Official Account draft box.

Project implementation and current boundaries

The interface uses Next.js, React, and a rich-text editor; business content is stored in Supabase. Research creation tasks are executed by an independent Worker and save task status with queues and workflows; whole-web retrieval, writing, and image generation each depend on configured external services. Before deployment and use, you need to configure the database, model, retrieval service, image service, and WeChat Official Account credentials according to the project documentation.

Currently, this is a project that has implemented the core creation pipeline, but creators still need to do final verification of sources, facts, and image text. The project documentation also clearly states that enterprise-level identity migration, complete evaluation, failure drills, and production monitoring have not all been completed. Therefore, this introduction describes the creation workflow that the existing workbench can complete, without writing it as an automatic publishing system that has passed all production-level acceptance.

What Content Studio hopes to save is the time spent repeatedly moving content between different tools. It lets topics, sources, articles, images, and platform materials gradually take shape along the same line, while leaving review and the final publishing decision to the creator.